pax_global_header 0000666 0000000 0000000 00000000064 13203113151 0014500 g ustar 00root root 0000000 0000000 52 comment=1512f11a18b70c61845b35d5a017004d1b20b1f5
python-casacore-2.2.1/ 0000775 0000000 0000000 00000000000 13203113151 0014601 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/.coveragerc 0000664 0000000 0000000 00000000134 13203113151 0016720 0 ustar 00root root 0000000 0000000 [report]
omit =
*/python?.?/*
*/site-packages/nose/*
*__init__*
*/tests/*
python-casacore-2.2.1/.dockerignore 0000664 0000000 0000000 00000000056 13203113151 0017256 0 ustar 00root root 0000000 0000000 .git
.venv*/
.virtualenv*/
build/
*.egg-info/
python-casacore-2.2.1/.gitignore 0000664 0000000 0000000 00000000235 13203113151 0016571 0 ustar 00root root 0000000 0000000 .idea
doc/.build/
build/
dist/
*.pyc
.vagrant/
casacore/*/_*.so
python_casacore.egg-info/
casacore.egg-info/
*~
.coverage
.virtualenv/
.virtualenv3/
htmlcov
python-casacore-2.2.1/.travis.yml 0000664 0000000 0000000 00000001156 13203113151 0016715 0 ustar 00root root 0000000 0000000 language: generic
sudo: required
dist: trusty
os:
- linux
- osx
matrix:
allow_failures:
- os: osx
before_install:
- .travis/before_install.sh
install:
- .travis/install.sh
script:
- .travis/script.sh
deploy:
provider: pypi
user: gijzelaerr
password:
secure: cdrrma3XaFjtv4lHvag6dfhSbKBF8iLmVgPQEjXP8qa+WxHnkHObkyraYAFGqThDY/0lBdrBm7Og6l1JfAcSA2BjdPQYxujP8KEoVicPwlgwEJ5LZo4HqympWVk33APvbcYNw7K/CwEXNJCCD1tDiO4pdwkPAWuKlnYUVfZq2yI=
on:
tags: true
repo: casacore/python-casacore
after_success:
coveralls
notifications:
webhooks:
urls:
- https://webhooks.gitter.im/e/08a570c12a3afa37d8e2
python-casacore-2.2.1/.travis/ 0000775 0000000 0000000 00000000000 13203113151 0016167 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/.travis/Dockerfile 0000664 0000000 0000000 00000001071 13203113151 0020160 0 ustar 00root root 0000000 0000000 FROM kernsuite/base:3
RUN apt-get update && apt-get install -y casacore-data casacore-dev \
libboost-python-dev libcasa-python3-2 libcfitsio3-dev python-dev \
python-numpy python-setuptools python-six python3-all python3-dev \
python3-numpy python3-setuptools python3-six wcslib-dev python-pip \
python3-pip python-nose python3-nose
ADD . /code
WORKDIR /code
RUN pip install -e .
RUN pip3 install -e .
RUN pip install -r tests/requirements.txt
RUN pip3 install -r tests/requirements.txt
RUN nosetests --with-coverage
RUN nosetests3 --with-coverage
python-casacore-2.2.1/.travis/before_install.sh 0000775 0000000 0000000 00000001513 13203113151 0021516 0 ustar 00root root 0000000 0000000 #!/usr/bin/env bash
set -e
set -v
if [ "$TRAVIS_OS_NAME" = linux ]; then
true
fi
if [ "$TRAVIS_OS_NAME" = osx ]; then
# isntall and configure miniconda
wget https://repo.continuum.io/miniconda/Miniconda2-latest-MacOSX-x86_64.sh -O miniconda.sh;
bash miniconda.sh -b -p $HOME/miniconda
export PATH="$HOME/miniconda/bin:$PATH"
hash -r
conda config --set always_yes yes --set changeps1 no
conda update -q conda
conda config --add channels conda-forge
# Useful for debugging any issues with conda
conda info -a
conda create -q -n testenv python=2.7 casacore=2.3.0
source activate testenv
# setup casacore data
mkdir $HOME/data
cd $HOME/data
wget ftp://ftp.astron.nl/outgoing/Measures/WSRT_Measures.ztar && tar xf WSRT_Measures.ztar &&
cd $HOME
echo "measures.directory: $HOME/data" > $HOME/.casarc
fi
python-casacore-2.2.1/.travis/install.sh 0000775 0000000 0000000 00000000526 13203113151 0020177 0 ustar 00root root 0000000 0000000 #!/usr/bin/env bash
set -e
set -v
if [ "$TRAVIS_OS_NAME" = linux ]; then
true
fi
if [ "$TRAVIS_OS_NAME" = osx ]; then
cd $TRAVIS_BUILD_DIR
python setup.py build_ext \
-L$HOME/miniconda/envs/testenv/lib \
-I$HOME/miniconda/envs/testenv/include
pip install -e .
pip install -r tests/requirements.txt
fi
python-casacore-2.2.1/.travis/script.sh 0000775 0000000 0000000 00000000571 13203113151 0020035 0 ustar 00root root 0000000 0000000 #!/usr/bin/env bash
set -e
set -v
if [ "$TRAVIS_OS_NAME" = linux ]; then
cd $TRAVIS_BUILD_DIR
docker build -f .travis/Dockerfile .
fi
if [ "$TRAVIS_OS_NAME" = osx ]; then
export DYLD_LIBRARY_PATH=$HOME/miniconda/envs/testenv/lib
export LD_LIBRARY_PATH=$HOME/miniconda/envs/testenv/lib
nosetests --with-coverage
travis-sphinx --nowarn -s doc build
fi
python-casacore-2.2.1/CHANGELOG.md 0000664 0000000 0000000 00000001503 13203113151 0016411 0 ustar 00root root 0000000 0000000 # 2.2.0
- Expose MeasurementSet functionality (#61)
- Many improvements to documentation and test coverage
Thanks to Shibasis Patel (@shibasisp) our Google summer of Code '17 student.
A full list of changes can be found on the issue tracker:
https://github.com/casacore/python-casacore/milestone/6?closed=1
# 2.1.0
- Replaced references to pyrap -> casacore (issue #16)
- Experimental support for Python3 (needs python3 branch build of casacore)
- Link against correct Python on OSX when using homebrew (issue #15)
# 2.0.0
- Renamed project from pyrap to python-casacore
- Renamed modules from pyrap to casacore
- Added backswards compatible module which imports casacore
- Removed scons build dependency, just use setup.py
- Depends on CASACORE 2.0
- Cleanup of project structure
- Moved development to github
python-casacore-2.2.1/LICENSE 0000664 0000000 0000000 00000043153 13203113151 0015614 0 ustar 00root root 0000000 0000000 GNU GENERAL PUBLIC LICENSE
Version 2, June 1991
Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The licenses for most software are designed to take away your
freedom to share and change it. By contrast, the GNU General Public
License is intended to guarantee your freedom to share and change free
software--to make sure the software is free for all its users. This
General Public License applies to most of the Free Software
Foundation's software and to any other program whose authors commit to
using it. (Some other Free Software Foundation software is covered by
the GNU Lesser General Public License instead.) You can apply it to
your programs, too.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
this service if you wish), that you receive source code or can get it
if you want it, that you can change the software or use pieces of it
in new free programs; and that you know you can do these things.
To protect your rights, we need to make restrictions that forbid
anyone to deny you these rights or to ask you to surrender the rights.
These restrictions translate to certain responsibilities for you if you
distribute copies of the software, or if you modify it.
For example, if you distribute copies of such a program, whether
gratis or for a fee, you must give the recipients all the rights that
you have. You must make sure that they, too, receive or can get the
source code. And you must show them these terms so they know their
rights.
We protect your rights with two steps: (1) copyright the software, and
(2) offer you this license which gives you legal permission to copy,
distribute and/or modify the software.
Also, for each author's protection and ours, we want to make certain
that everyone understands that there is no warranty for this free
software. If the software is modified by someone else and passed on, we
want its recipients to know that what they have is not the original, so
that any problems introduced by others will not reflect on the original
authors' reputations.
Finally, any free program is threatened constantly by software
patents. We wish to avoid the danger that redistributors of a free
program will individually obtain patent licenses, in effect making the
program proprietary. To prevent this, we have made it clear that any
patent must be licensed for everyone's free use or not licensed at all.
The precise terms and conditions for copying, distribution and
modification follow.
GNU GENERAL PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
0. This License applies to any program or other work which contains
a notice placed by the copyright holder saying it may be distributed
under the terms of this General Public License. The "Program", below,
refers to any such program or work, and a "work based on the Program"
means either the Program or any derivative work under copyright law:
that is to say, a work containing the Program or a portion of it,
either verbatim or with modifications and/or translated into another
language. (Hereinafter, translation is included without limitation in
the term "modification".) Each licensee is addressed as "you".
Activities other than copying, distribution and modification are not
covered by this License; they are outside its scope. The act of
running the Program is not restricted, and the output from the Program
is covered only if its contents constitute a work based on the
Program (independent of having been made by running the Program).
Whether that is true depends on what the Program does.
1. You may copy and distribute verbatim copies of the Program's
source code as you receive it, in any medium, provided that you
conspicuously and appropriately publish on each copy an appropriate
copyright notice and disclaimer of warranty; keep intact all the
notices that refer to this License and to the absence of any warranty;
and give any other recipients of the Program a copy of this License
along with the Program.
You may charge a fee for the physical act of transferring a copy, and
you may at your option offer warranty protection in exchange for a fee.
2. You may modify your copy or copies of the Program or any portion
of it, thus forming a work based on the Program, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:
a) You must cause the modified files to carry prominent notices
stating that you changed the files and the date of any change.
b) You must cause any work that you distribute or publish, that in
whole or in part contains or is derived from the Program or any
part thereof, to be licensed as a whole at no charge to all third
parties under the terms of this License.
c) If the modified program normally reads commands interactively
when run, you must cause it, when started running for such
interactive use in the most ordinary way, to print or display an
announcement including an appropriate copyright notice and a
notice that there is no warranty (or else, saying that you provide
a warranty) and that users may redistribute the program under
these conditions, and telling the user how to view a copy of this
License. (Exception: if the Program itself is interactive but
does not normally print such an announcement, your work based on
the Program is not required to print an announcement.)
These requirements apply to the modified work as a whole. If
identifiable sections of that work are not derived from the Program,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works. But when you
distribute the same sections as part of a whole which is a work based
on the Program, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote it.
Thus, it is not the intent of this section to claim rights or contest
your rights to work written entirely by you; rather, the intent is to
exercise the right to control the distribution of derivative or
collective works based on the Program.
In addition, mere aggregation of another work not based on the Program
with the Program (or with a work based on the Program) on a volume of
a storage or distribution medium does not bring the other work under
the scope of this License.
3. You may copy and distribute the Program (or a work based on it,
under Section 2) in object code or executable form under the terms of
Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable
source code, which must be distributed under the terms of Sections
1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three
years, to give any third party, for a charge no more than your
cost of physically performing source distribution, a complete
machine-readable copy of the corresponding source code, to be
distributed under the terms of Sections 1 and 2 above on a medium
customarily used for software interchange; or,
c) Accompany it with the information you received as to the offer
to distribute corresponding source code. (This alternative is
allowed only for noncommercial distribution and only if you
received the program in object code or executable form with such
an offer, in accord with Subsection b above.)
The source code for a work means the preferred form of the work for
making modifications to it. For an executable work, complete source
code means all the source code for all modules it contains, plus any
associated interface definition files, plus the scripts used to
control compilation and installation of the executable. However, as a
special exception, the source code distributed need not include
anything that is normally distributed (in either source or binary
form) with the major components (compiler, kernel, and so on) of the
operating system on which the executable runs, unless that component
itself accompanies the executable.
If distribution of executable or object code is made by offering
access to copy from a designated place, then offering equivalent
access to copy the source code from the same place counts as
distribution of the source code, even though third parties are not
compelled to copy the source along with the object code.
4. You may not copy, modify, sublicense, or distribute the Program
except as expressly provided under this License. Any attempt
otherwise to copy, modify, sublicense or distribute the Program is
void, and will automatically terminate your rights under this License.
However, parties who have received copies, or rights, from you under
this License will not have their licenses terminated so long as such
parties remain in full compliance.
5. You are not required to accept this License, since you have not
signed it. However, nothing else grants you permission to modify or
distribute the Program or its derivative works. These actions are
prohibited by law if you do not accept this License. Therefore, by
modifying or distributing the Program (or any work based on the
Program), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Program or works based on it.
6. Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties to
this License.
7. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Program at all. For example, if a patent
license would not permit royalty-free redistribution of the Program by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Program.
If any portion of this section is held invalid or unenforceable under
any particular circumstance, the balance of the section is intended to
apply and the section as a whole is intended to apply in other
circumstances.
It is not the purpose of this section to induce you to infringe any
patents or other property right claims or to contest validity of any
such claims; this section has the sole purpose of protecting the
integrity of the free software distribution system, which is
implemented by public license practices. Many people have made
generous contributions to the wide range of software distributed
through that system in reliance on consistent application of that
system; it is up to the author/donor to decide if he or she is willing
to distribute software through any other system and a licensee cannot
impose that choice.
This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.
8. If the distribution and/or use of the Program is restricted in
certain countries either by patents or by copyrighted interfaces, the
original copyright holder who places the Program under this License
may add an explicit geographical distribution limitation excluding
those countries, so that distribution is permitted only in or among
countries not thus excluded. In such case, this License incorporates
the limitation as if written in the body of this License.
9. The Free Software Foundation may publish revised and/or new versions
of the General Public License from time to time. Such new versions will
be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the Program
specifies a version number of this License which applies to it and "any
later version", you have the option of following the terms and conditions
either of that version or of any later version published by the Free
Software Foundation. If the Program does not specify a version number of
this License, you may choose any version ever published by the Free Software
Foundation.
10. If you wish to incorporate parts of the Program into other free
programs whose distribution conditions are different, write to the author
to ask for permission. For software which is copyrighted by the Free
Software Foundation, write to the Free Software Foundation; we sometimes
make exceptions for this. Our decision will be guided by the two goals
of preserving the free status of all derivatives of our free software and
of promoting the sharing and reuse of software generally.
NO WARRANTY
11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
REPAIR OR CORRECTION.
12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
{description}
Copyright (C) {year} {fullname}
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
Also add information on how to contact you by electronic and paper mail.
If the program is interactive, make it output a short notice like this
when it starts in an interactive mode:
Gnomovision version 69, Copyright (C) year name of author
Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
This is free software, and you are welcome to redistribute it
under certain conditions; type `show c' for details.
The hypothetical commands `show w' and `show c' should show the appropriate
parts of the General Public License. Of course, the commands you use may
be called something other than `show w' and `show c'; they could even be
mouse-clicks or menu items--whatever suits your program.
You should also get your employer (if you work as a programmer) or your
school, if any, to sign a "copyright disclaimer" for the program, if
necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the program
`Gnomovision' (which makes passes at compilers) written by James Hacker.
{signature of Ty Coon}, 1 April 1989
Ty Coon, President of Vice
This General Public License does not permit incorporating your program into
proprietary programs. If your program is a subroutine library, you may
consider it more useful to permit linking proprietary applications with the
library. If this is what you want to do, use the GNU Lesser General
Public License instead of this License.
python-casacore-2.2.1/MANIFEST.in 0000664 0000000 0000000 00000000067 13203113151 0016342 0 ustar 00root root 0000000 0000000 include README.rst *.py
recursive-include src *.cc *.h
python-casacore-2.2.1/README.rst 0000664 0000000 0000000 00000005056 13203113151 0016276 0 ustar 00root root 0000000 0000000 python-casacore
===============
Python-casacore is a set of Python bindings for `casacore `_,
a c++ library used in radio astronomy. Python-casacore replaces the old
`pyrap `_.
The python-casacore documentation can be found on `casacore.github.io/python-casacore `_.
.. image:: https://travis-ci.org/casacore/python-casacore.svg?branch=master
:target: https://travis-ci.org/casacore/python-casacore
.. image:: https://coveralls.io/repos/github/casacore/python-casacore/badge.svg?branch=master
:target: https://coveralls.io/github/casacore/python-casacore?branch=master
Installation
============
Ubuntu 16.04
------------
The easiest way to start using python-casacore is to enable the `KERN suite `_ repository and install the binary package ::
$ sudo apt-get install software-properties-common
$ sudo add-apt-repository ppa:kernsuite/kern-3
$ sudo apt-get update
$ sudo apt-get install python-casacore
from source
-----------
install these requirements:
* `setuptools `_
* `Casacore `_ >= 2.3
* `Boost-python `_
* `numpy `_
* `cfitsio `_
On ubuntu you can install these with:
* enable the `KERN suite `_
* install build dependencies::
$ apt-get install casacore-dev python-numpy \
python-setuptools libboost-python-dev libcfitsio3-dev wcslib-dev
* compile and install::
$ python ./setup.py install
* if you need to supply compile parameters, for example if you have your casacore
library installed in a different location have a look at the `build_ext` help::
$ python ./setup.py build_ext -h
For example, if `casacore` is installed in `/opt/local/`, you can specify the
library path and include path with::
$ python ./setup.py build_ext -I/opt/local/include -L/opt/local/lib
$ python ./setup.py install
Using pip
---------
python-casacore is also available trough pip. Note that you need to manually satisfy
the requirements mentioned above::
$ pip install python-casacore
Support
=======
if you have any problems, suggestions or questions please open an issue on the
python-casacore github issue tracker.
Credits
=======
* `Ger van Diepen `_
* `Malte Marquarding `_
* `Gijs Molenaar `_
python-casacore-2.2.1/Vagrantfile 0000664 0000000 0000000 00000000551 13203113151 0016767 0 ustar 00root root 0000000 0000000 # -*- mode: ruby -*-
# vi: set ft=ruby :
VAGRANTFILE_API_VERSION = "2"
Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|
config.vm.box = "ubuntu/trusty64"
config.vm.provision :shell, :path => "vagrant/provision.sh"
# default vagrant config is not really enough
config.vm.provider "virtualbox" do |v|
v.memory = 2048
v.cpus = 2
end
end
python-casacore-2.2.1/casacore/ 0000775 0000000 0000000 00000000000 13203113151 0016361 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/__init__.py 0000664 0000000 0000000 00000000067 13203113151 0020475 0 ustar 00root root 0000000 0000000 __version__ = "2.2.1"
__mincasacoreversion__ = "2.3.0"
python-casacore-2.2.1/casacore/fitting/ 0000775 0000000 0000000 00000000000 13203113151 0020025 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/fitting/__init__.py 0000775 0000000 0000000 00000034453 13203113151 0022152 0 ustar 00root root 0000000 0000000 # __init__.py: Top level .py file for python ftting interface
# Copyright (C) 2006,2007
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: __init__.py,v 1.1 2006/10/20 06:30:03 mmarquar Exp $
"""Python interface to the Casacore scimath fitting module.
Introduction
============
The fitting module provides least squares fitting. It can handle linear and
non-linear; real and complex (including cases where unknowns are each other's
conjugate); complete and singular-value-decomposition; with or without
external constraints; general or specific cases.
------------------
The fitting module
------------------
For most uses we will create a single fitting object to work with::
from casacore.fitting import fitserver
dfit = fitserver()
More fitting tools can be created by either the fitter constructor (which
creates and returns a separate fitting tool), or by the fitter method of
an existing fitting tool, which returns a fit identifier, which can be
used to indicate a specific sub-fitter in the fitter used by including a
parameter 'id=' in all calls to the fitting tool's functions. The latter is
especially useful in the case where many simultaneous solutions are necessary:
it is more resource efficient, and also allows you to have an array of fit
indices to loop over. In both cases the parameters of the tool can be given
in the constructor (fitter method), or in a separate init method (see next
example of the highest level use)::
from casacore.fitting import fitserver
myfit = fitserver() # general fitting object created
# (needs initializing before it can be used)
cpid = myfit.fitter(ftype='complex') # and another (sub-)fitter
# with an id
The theory behind the fitting module's operation is described in detail in
(`Note 224 <../../casacore/doc/notes/224.html>`_).
Fitting requires a model describing the data obtained. The model is a described
as a functional with parameters to be solved for. Functionals can be
pre-programmed functionals like poly, gauss1d, or free form like compiled. In
the latter case an expression string describes the model.
The model can depend on zero, one or more arguments, called x. The number of
arguments determines the dimension of the model.
Fitting also needs a set of data, called y. If the model is not 0-dimensional,
each value x will have an observed value. E.g. for each hour of the day x you
can have a measured temperature. Or in the case of multi-dimensions e.g. each
pair of hour of the day at each height above the surface (x0, x1) you should
have a y value.
--------
Examples
--------
Simple linear example
---------------------
The example uses a set of x coordinates::
from numpy import arange
x = -1 + 0.1*arange(21)
The 'observed' values used are a simple 1-dim polynomial of order 2:
1 + 2(x+1) + 0.03(x+1)^2 == 3.03 +2.06x + 0.03x^2
We fill these values using the polynomial functional::
y = functionals.poly(2, [3.03, 2.06, 0.03])(x)
To take the average of these points, we can do::
dfit.linear(functionals.compiled('p'), [], y)
Note that an expression uses p (which is p0), p1 and x, x0, x1. Note also that
since no argument is used in the expression, no x-values have to be given.
We can get solutions and errors from these data (see for details the separate
routine descriptions)::
>>> dfit.solution()
3.041
# Compare with the result:
>>> print sum(y)/len(y)
3.041
>>> print dfit.sd() # standard deviation per observation
1.27824
>>> print dfit.stddev() # standard deviation per weight
1.27824
>>> print dfit.error() # errors in solved parameters
array( [0.278934])
>>> print dfit.rank() # rank of solution
1
>>> print dfit.covariance() # covariance matrix
[[0.047619]]
We can also try to use a 0-order polynomial. Note that a polynomial, even a
zero-order one, is a 1-dim function, and we need an x defined::
>>> dfit.linear(dfs.poly(0), [], y)
RuntimeError: Linear fitter x and y lengths disagree
>>> dfit.linear(dfs.poly(0),x,y)
print >>> dfit.solution()
3.041
We would like to check the results, so we will do an average in a separate
fitter::
>>> id = dfit.fitter() # get a new fitter
>>> dfit.linear(dfs.compiled('p'), [], y, fid=id) # get average
>>> dfit.solution() - dfit.solution(fid=id) # check difference
-4.44089e-16
# to really show we recalculate and check separately:
>>> dfit.linear(dfs.compiled('p'), [], array(y)/2, fid=id) # calculate new average
>>> print dfit.solution() - dfit.solution(fid=id)
[1 .5205]
A 1-order polynomial is now easy::
>>> dfit.linear(dfs.poly(1), x, y)
>>> print dfit.solution()
[3.041, 2.06]
>>> print dfit.chi2()
0.00201894
>>> print dfit.error()
[0.00224944, 0.00371484]
>>> print dfit.sd()
0.0103082
Note that each 'equation' can also be given a weight or standard deviation.
2-dimensional example
---------------------
A 2-dim model is done the same way. The x vector has now n pairs of values.
The Glish rbind can help in creating these pairs::
>>> x1 = arange(1, 6)
>>> x2 = 0.1*x1
>>> x1 = ravel(array([x1,x2]).transpose()) # combine into pairs. Check:
>>> print x1
array([ 1. , 0.1, 2. , 0.2, 3. , 0.3, 4. , 0.4, 5. , 0.5])
>>> dfit.linear(dfs.compiled('p*x + p1*sin(x1)'), x1,
dfs.compiled('3*x+7*sin(x[2])').f(x1))
>>> print dfit.solution()
[ 3. 7.]
Non-linear simple example
-------------------------
If the model is non-linear in the parameters to be solved, the functional
method should be used. The main difference is that a guess solution must be
inserted in the model parameters. In the following that is not necessary,
since the default zero values suffice if the function is linear::
>>> dfit.functional(dfs.compiled('p*x + p1*sin(x1)'), x1,
dfs.compiled('3*x+7*sin(x[2])').f(x1), id=id)
>>> dfit.solution(fid=id)
[ 3. 7.]
>>> dfit.solution(fid=id)-dfit.solution()
[ 6.17284002e-13 -6.35846931e-12]
# Try with an intial guess
>>> dfit.functional(dfs.compiled('p*x + p1*sin(x1)', [3,7]), x1,
dfs.compiled('3*x+7*sin(x[2])').f(x1), fid=id2)
>>> dfit.solution(fid=id2)
[ 3. 7.]
>>> dfit.solution(fid=id2)-dfit.solution()
[ 6.17284002e-13 -6.35846931e-12]
Functional variety
------------------
Just to show the model can be anything, we redo the fit of an order 1
polynomial to the x, y data::
>>> dfit.linear(dfs.poly(1), x,y)
>>> dfit.solution()
[ 3.041 2.06]
Now try the same by a sum of odd and even polynomials of default order
(note the order)::
a = dfs.compound()
a.add(dfs.functional('oddp'))
a.add(dfs.functional('evenp'))
dfit.linear(a,x,y,id=id2)
dfit.solution(id=id2)
[ 2.06 3.041]
And the combination of an odd (2x) and an even polynomial (3)::
a = dfs.combi()
a.add(dfs.functional('oddp', params=2))
a.add(dfs.functional('evenp', params=3))
dfit.linear(a, x, y)
>>> dfit.solution()
[ 1.03 1.01367]
Use constraints
---------------
We have measured a number of anlgles around a triangle. Each angle is measured
10 times (nominally 50, 60, 70 deg). Solving the angles will give::
>>> import numpy
>>> yz = numpy.array([numpy.zeros(10) + 50 + numpy.random.normal(0,1,10),
numpy.zeros(10) + 60 + numpy.random.normal(0,1,10),
numpy.zeros(10) + 70 + numpy.random.normal(0,1,10)]).flatten()
# Create 3*10 equations
>>> xz = array([1,0,0]*10 + [0,1,0]*10 + [0,0,1]*10)
# The equation used and solve
>>> f = dfs.compiled('p*x+p1*x1+p2*x2')
>>> dfit.linear(f, xz, yz)
>>> print dfit.solution(), 'sum=', sum(dfit.solution())
[49.7079 60.2427 70.092] sum= 180.043
>>> dfit.error()
[ 0.334828 0.334828 0.334828]
# Add a constraint: sum of angles 180deg
>>> dfit.addconstraint(x=[1,1,1],y=180)
>>> dfit.linear(f,xz,yz)
>>> print dfit.solution(), 'sum=', sum(dfit.solution())
[ 49.6937 60.2285 70.0778] sum= 180
>>> print dfit.error()
[ 0.273413 0.273413 0.273413]
# Add another constraint, since we know second angle 60deg
>>> dfit.addconstraint(x=[0,1,0], y=60)
>>> dfit.linear(f,xz,yz)
>>> print dfit.solution(), 'sum=', sum(dfit.solution())
[ 49.8079 60 70.1921] sum= 180
>>> print dfit.error()
[0.239827 0 0.239827]
Non-linear equation and constraints
-----------------------------------
In the following we have 2 Gaussian profiles and an offset. We add some noise,
and solve assuming we have a fair estimate of the position of the Gaussians.
Note that if the first estimate is beyond the real half-value point, the
fitting will be difficult, due to the derivatives changing sign::
# The profile to generate and the parameters to use
# (in essence 10 + 20 * exp (-((x-10)/4)^2) + 10 * exp(-((x-33)/4)^2) )
f = dfs.compiled('p6+p0*exp(-((x-p1)/p2)^2) + p3*exp(-((x-p4)/p5)^2)',
[20, 10, 4, 10, 33, 4, 10])
xg = 0.5 * numpy.arange(1, 101) - 0.5
yg = numpy.array(f(xg)) + numpy.random.normal(0,0.3,100)
# Make an intial guess
f.set_parameters([22, 11, 5, 10, 30, 5, 9])
# Solve
dfit.clearconstraints()
dfit.functional(f,xg,yg)
print dfit.solution()
print dfit.solution() - numpy.array([20., 10, 4, 10, 33, 4, 10])
print dfit.error()
[0.211312 0.0334257 0.0527771 0.213666 0.0652003 0.102782 0.082641]
# We know that the two lines have a peak ratio of 2: Amp1-2Amp2 = 0
dfit.addconstraint([1, 0, 0, -2, 0, 0, 0])
dfit.functional(f, xg, yg)
print dfit.solution()
print dfit.solution() - numpy.array([20., 10, 4, 10, 33, 4, 10])
print dfit.solution()[0]/dfit.solution()[3]
print dfit.error()
# We know that the lines originated in same place: width1 == width2
# Note that the default assumed value is 0.0
dfit.addconstraint([0, 0, 1, 0, 0, -1, 0])
dfit.functional(f, xg, yg)
print dfit.solution()
dfit.solution() - numpy.array([20, 10, 4, 10, 33, 4, 10])
dfit.solution()[2]-dfit.solution()[5]
dfit.error()
# And see what happens if we assume that the widths are 4
dfit.addconstraint([0, 0, 1, 0, 0, 0, 0], 4)
dfit.functional(f, xg, yg)
dfit.solution()
dfit.solution() - [20, 10, 4, 10, 33, 4, 10]
dfit.error()
Deficient solutions and SVD constraints
---------------------------------------
*DOES NOT WORK*
In some cases solutions of the least-squares equations is not completely
possible. An example is e.g. the solution of the closures equations in
synthesis calibrations, where a missing phase zero and slope and a missing
absolute gain cannot be solved for. The fitting described here will always
provide a solution, even in the case of a set of incomplete equations. After
the solution the deficiency can be checked. If there is a rank deficiency,
the set of 'constraints' that makes a solution possible (in a way similar to
SVD, i.e. providing a missing set of orthogonal equations) is available
through the constr function::
# Provide a set of equations.
x = array([1,1,1]*10)
y = 180 + numpy.zeros(10) + numpy.random.normal(0, 3, 10)
f = dfs.functional('hyper', 3)
dfit.linear(f,x,y)
dfit.deficiency()
2
dfit.solution()
[60.0262 60.0262 60.0262]
dfit.constraint()
[-1 0 1 -1 1 0]
# The SVD constraints can be used as constraints in subsequent solutions:
dfit.addconstraint(x=dfit.constraint(1))
T
dfit.addconstraint(x=dfit.constraint(2))
T
dfit.linear(f,xz,yz)
T
dfit.solution()
[60.0262 60.0262 60.0262]
dfit.rank()
3
dfit.deficiency()
0
dfit.error()
[0.202801 0.202801 0.202801]
Complex fitting
---------------
The fitter can handle functions of complex variables. In the following example
a second order polynomial is first fitted real with a first order linear
polynomial. The same is repeated complex (with real data); and then a complex
value is fitted. An example of a 2-dimensional non-linear function is also
given::
# Define x and y data
>>> x = -1 + numpy.arange(0,21)*0.1
>>> y = dfs.poly(2, [3.03, 2.06, 0.03])(x)
# fit a first order polynomial
>>> dfit.linear(dfs.poly(1), x,y)
>>> print 'linear', dfit.solution()
linear [ 3.041 2.06]
# Get a complex fitter and see the same fit
>>> id1 = dfit.fitter()
>>> dfit.set(ftype='complex', fid=id1)
>>> dfit.linear(dfs.poly(1, dtype='complex'), x, y, fid=id1)
>>> dfit.solution(fid=id1)
[ 3.041+0j 2.06+0j]
# Make a complex yi and redo
>>> yi = dfs.poly(2, [3.03, 2.06, 0.03])(x)
>>> yi = yi - 3j*array(dfs.poly(2, [3.03+0j, 2.06, 0.03])(x))
>>> dfit.linear(dfs.poly(1, dtype='complex'), x, yi, fid=id1)
>>> dfit.solution(fid=id1)
[ 3.041-9.123j 2.06-6.18j]
# A non-linear 2-dimensional function, real and complex
>>> id2 = dfit.fitter()
>>> dfit.functional(dfs.compiled('p*x + p1*sin(x1)', [3,7]), x1,
dfs.compiled('3*x+7*sin(x[2])')(x1), fid=id2)
>>> x1 = arange(1,6)
>>> x2 = 0.1*x1
>>> x1 = array(zip(x1,x2)).flatten()
>>> dfit.functional(dfs.compiled('p*x + p1*sin(x1)', [3,7]),x1,
dfs.compiled('3*x+7*sin(x[2])').f(x1), fid=id2)
>>> dfit.solution(fid=id2)
>>> dfit.set(type=dfit.complex(), fid=id2)
>>> dfit.functional(dfs.compiled('p*x + p1*sin(x1)', [3,7]),x1,
dfs.compiled('3*x+7*sin(x[2])').f(x1), fid=id2)
>>> dfit.solution(fid=id2)
"""
from .fitting import *
python-casacore-2.2.1/casacore/fitting/fitting.py 0000664 0000000 0000000 00000042626 13203113151 0022055 0 ustar 00root root 0000000 0000000 from ._fitting import fitting
import numpy as NUM
from casacore import six
from casacore.functionals import *
class fitserver(object):
"""Create a `fitserver instance.
The object can be created without arguments (in which case it is assumed
to be a real fitter), or with the arguments specifying the number of
unknowns to be solved for (a number not relevant in practice); and the type
of solution: real, complex, conjugate (complex with both the unknown and
its conjugate in the condition equations), separable complex, asreal
complex with the real and imaginary part seen as independent unknowns.
All solutions need a model (specified as a :mod:`casacore.functionals`.
All solutions are done using an SVD type method. A collinearity factor
can be specified, which is in essence the sine squared of the minimum angle
between two normal equation columns that are still to be considered
independent. For automatic non-linear solutions, a Levenberg-Marquardt
factor (see`Note 224 <../../casacore/doc/notes/224.html>`_) is used, which
can be specified as well.
In the case of non-linear solutions that have to be handled by the system,
an initial estimate for the model parameters is necessary.
:param n: number of unknowns
:param ftype: type of solution
Allowed: real, complex, separable, asreal, conjugate
:param colfac: collinearity factor
:param lmfac: Levenberg-Marquardt factor
:param fid: the id of a sub-fitter
"""
def __init__(self, n=0, m=1, ftype=0, colfac=1.0e-8, lmfac=1.0e-3):
self._fitids = []
self._typeids = {"real": 0, "complex": 1, "separable": 3,
"asreal": 7, "conjugate": 11}
self._fitproxy = fitting()
fid = self.fitter(n=n, ftype=ftype, colfac=colfac, lmfac=lmfac)
if fid != 0:
raise RuntimeError("System problem creating fitter server")
def fitter(self, n=0, ftype="real", colfac=1.0e-8, lmfac=1.0e-3):
"""Create a sub-fitter.
The created sub-fitter can be used in the same way as a fitter
default fitter. This function returns an identification, which has to
be used in the `fid` argument of subsequent calls. The call can
specify the standard constructor arguments (`n`, `type`, `colfac`,
`lmfac`), or can specify them later in a :meth:`set` statement.
:param n: number of unknowns
:param ftype: type of solution
Allowed: real, complex, separable, asreal, conjugate
:param colfac: collinearity factor
:param lmfac: Levenberg-Marquardt factor
:param fid: the id of a sub-fitter
"""
fid = self._fitproxy.getid()
ftype = self._gettype(ftype)
n = len(self._fitids)
if 0 <= fid < n:
self._fitids[fid] = {}
elif fid == n:
self._fitids.append({})
else:
# shouldn't happen
raise RangeError("fit id out of range")
self.init(n=n, ftype=ftype, colfac=colfac, lmfac=lmfac, fid=fid)
return fid
def init(self, n=0, ftype="real", colfac=1.0e-8, lmfac=1.0e-3, fid=0):
"""Set selected properties of the fitserver instance.
Like in the constructor, the number of unknowns to be solved for;
the number of simultaneous solutions; the ftype and the collinearity
and Levenberg-Marquardt factor can be specified. Individual values can
be overwritten with the :meth:`set` function.
:param n: number of unknowns
:param ftype: type of solution
Allowed: real, complex, separable, asreal, conjugate
:param colfac: collinearity factor
:param lmfac: Levenberg-Marquardt factor
:param fid: the id of a sub-fitter
"""
ftype = self._gettype(ftype)
self._fitids[fid]["stat"] = False
self._fitids[fid]["solved"] = False
self._fitids[fid]["haserr"] = False
self._fitids[fid]["fit"] = False
self._fitids[fid]["looped"] = False
if self._fitproxy.init(fid, n, ftype, colfac, lmfac):
self._fitids[fid]["stat"] = self._getstate(fid)
else:
return False
def _gettype(self, ftype):
if isinstance(ftype, str):
ftype = ftype.lower()
if ftype not in self._typeids:
raise TypeError("Illegal fitting type")
else:
return self._typeids[ftype]
elif isinstance(ftype, int):
if ftype not in self._typeids.values():
raise TypeError("Illegal fitting type")
else:
raise TypeError("Illegal fitting type")
return ftype
def _settype(self, ftype=0):
for k, v in self._typeids.items():
if ftype == v:
return k
return "real"
def _checkid(self, fid=0):
if not (0 <= fid < len(self._fitids)
and isinstance(self._fitids[fid], dict)
and "stat" in self._fitids[fid]
and isinstance(self._fitids[fid]["stat"], dict)):
raise ValueError("fit id out of range")
def _reshape(self, fid=0):
pass
def _getstate(self, fid):
d = self._fitproxy.getstate(fid)
if "typ" in d:
d["typ"] = self._settype(d["typ"])
return d
def set(self, n=None, ftype=None, colfac=None, lmfac=None, fid=0):
"""Set selected properties of the fitserver instance.
All unset properties remain the same (in the :meth:`init` method all
properties are (re-)initialized). Like in the constructor, the number
of unknowns to be solved for; the number of simultaneous solutions;
the ftype (as code); and the collinearity and Levenberg-Marquardt
factor can be specified.
:param n: number of unknowns
:param ftype: type of solution
Allowed: real, complex, separable, asreal, conjugate
:param colfac: collinearity factor
:param lmfac: Levenberg-Marquardt factor
:param fid: the id of a sub-fitter
"""
self._checkid(fid)
if ftype is None:
ftype = -1
else:
ftype = self._gettype(ftype)
if n is None:
n = -1
elif n < 0:
raise ValueError("Illegal set argument n")
if colfac is None:
colfac = -1
elif colfac < 0:
raise ValueError("Illegal set argument colfac")
if lmfac is None:
lmfac = -1
elif lmfac < 0:
raise ValueError("Illegal set argument lmfac")
self._fitids[fid]["stat"] = False
self._fitids[fid]["solved"] = False
self._fitids[fid]["haserr"] = False
self._fitids[fid]["fit"] = True
self._fitids[fid]["looped"] = False
if n != -1 or ftype != -1 or colfac != -1 or lmfac != -1:
if not self._fitproxy.set(fid, n, ftype, colfac, lmfac):
return False
self._fitids[fid]["stat"] = self._getstate(fid)
return True
def done(self, fid=0):
"""Terminates the fitserver."""
self._checkid(fid)
self._fitids[fid] = {}
self._fitproxy.done(fid)
def reset(self, fid=0):
"""Reset the object's resources to its initialized state.
:param fid: the id of a sub-fitter
"""
self._checkid(fid)
self._fitids[fid]["solved"] = False
self._fitids[fid]["haserr"] = False
if not self._fitids[fid]["looped"]:
return self._fitproxy.reset(fid)
else:
self._fitids[fid]["looped"] = False
return True
def getstate(self, fid=0):
"""Obtain the state of the fitter object or a sub-fitter.
:param fid: the id of a sub-fitter
"""
self._checkid(fid)
return self._fitids[fid]["stat"]
def clearconstraints(self, fid=0):
"""Clear the constraints."""
self._checkid(fid)
self._fitids[fid]["constraint"] = {}
def addconstraint(self, x, y=0, fnct=None, fid=0):
"""Add constraint."""
self._checkid(fid)
i = 0
if "constraint" in self._fitids[fid]:
i = len(self._fitids[fid]["constraint"])
else:
self._fitids[fid]["constraint"] = {}
# dict key needs to be string
i = str(i)
self._fitids[fid]["constraint"][i] = {}
if isinstance(fnct, functional):
self._fitids[fid]["constraint"][i]["fnct"] = fnct.todict()
else:
self._fitids[fid]["constraint"][i]["fnct"] = \
functional("hyper", len(x)).todict()
self._fitids[fid]["constraint"][i]["x"] = [float(v) for v in x]
self._fitids[fid]["constraint"][i]["y"] = float(y)
six.print_(self._fitids[fid]["constraint"])
def fitpoly(self, n, x, y, sd=None, wt=1.0, fid=0):
if self.set(n=n + 1, fid=fid):
return self.linear(poly(n), x, y, sd, wt, fid)
def fitspoly(self, n, x, y, sd=None, wt=1.0, fid=0):
"""Create normal equations from the specified condition equations, and
solve the resulting normal equations. It is in essence a combination.
The method expects that the properties of the fitter to be used have
been initialized or set (like the number of simultaneous solutions m
the type; factors). The main reason is to limit the number of
parameters on the one hand, and on the other hand not to depend
on the actual array structure to get the variables and type. Before
fitting the x-range is normalized to values less than 1 to cater for
large difference in x raised to large powers. Later a shift to make x
around zero will be added as well.
:param n: the order of the polynomial to solve for
:param x: the abscissa values
:param y: the ordinate values
:param sd: standard deviation of equations (one or more values used
cyclically)
:param wt: an optional alternate for `sd`
:param fid: the id of the sub-fitter (numerical)
Example::
fit = fitserver()
x = N.arange(1,11) # we have values at 10 'x' values
y = 2. + 0.5*x - 0.1*x**2 # which are 2 +0.5x -0.1x^2
fit.fitspoly(3, x, y) # fit a 3-degree polynomial
print fit.solution(), fit.error() # show solution and their errors
"""
a = max(abs(max(x)), abs(min(x)))
if a == 0:
a = 1
a = 1.0 / a
b = NUM.power(a, range(n + 1))
if self.set(n=n + 1, fid=fid):
self.linear(poly(n), x * a, y, sd, wt, fid)
self._fitids[fid]["sol"] *= b
self._fitids[fid]["error"] *= b
return self.linear(poly(n), x, y, sd, wt, fid)
def fitavg(self, y, sd=None, wt=1.0, fid=0):
if self.set(n=1, fid=fid):
return self.linear(compiled("p"), [], y, sd, wt, fid)
def _fit(self, **kw):
fitfunc = kw.pop("fitfunc")
sd = kw.pop("sd")
fid = kw.pop("fid")
kw["id"] = fid
if not isinstance(kw["fnct"], functional):
raise TypeError("No or illegal functional")
if not self.set(n=kw["fnct"].npar(), fid=fid):
raise ValueError("Illegal fit id")
fnct = kw["fnct"]
kw["fnct"] = fnct.todict()
self.reset(fid)
x = self._as_array(kw["x"])
if x.ndim > 1 and fnct.ndim() == x.ndim:
x = x.flatten()
y = self._as_array(kw["y"])
if y.ndim > 1 and fnct.ndim() == y.ndim:
x = y.flatten()
wt = self._as_array(kw["wt"])
if sd is not None:
sd = self._as_array(sd)
wt = sd.copy()
wt[sd == 0] = 1
wt = 1 / abs(wt * NUM.conjugate(wt))
wt[NUM.logical_or(sd == -1, sd == 0)] = 0
ftype = fitfunc
dtype = 'float'
if (
self.getstate(fid)["typ"] != "real" or
NUM.iscomplexobj(x) or
NUM.iscomplexobj(y) or
NUM.iscomplexobj(wt)
):
ftype = "cx%s" % fitfunc
dtype = 'complex'
kw["x"] = self._as_array(x, dtype)
kw["y"] = self._as_array(y, dtype)
kw["wt"] = self._as_array(wt, dtype)
if "constraint" not in self._fitids[fid]:
self._fitids[fid]["constraint"] = {}
kw["constraint"] = self._fitids[fid]["constraint"]
func = getattr(self._fitproxy, ftype)
result = func(**kw)
self._fitids[fid].update(result)
self._fitids[fid]["solved"] = True
self._fitids[fid]["haserr"] = True
self._fitids[fid]["looped"] = False
def functional(self, fnct, x, y, sd=None, wt=1.0, mxit=50, fid=0):
"""Make a non-linear least squares solution.
This will make a non-linear least squares solution for the points
through the ordinates at the abscissa values, using the specified
`fnct`. Details can be found in the :meth:`linear` description.
:param fnct: the functional to fit
:param x: the abscissa values
:param y: the ordinate values
:param sd: standard deviation of equations (one or more values used
cyclically)
:param wt: an optional alternate for `sd`
:param mxit: the maximum number of iterations
:param fid: the id of the sub-fitter (numerical)
"""
self._fit(fitfunc="functional", fnct=fnct, x=x, y=y, sd=sd, wt=wt,
mxit=mxit, fid=fid)
nonlinear = functional
def linear(self, fnct, x, y, sd=None, wt=1.0, fid=0):
"""Make a linear least squares solution.
Makes a linear least squares solution for the points through the
ordinates at the x values, using the specified fnct. The x can be of
any dimension, depending on the number of arguments needed in the
functional evaluation. The values should be given in the order:
x0[1], x0[2], ..., x1[1], ..., xn[m] if there are n observations,
and m arguments. x should be a vector of m*n length; y (the
observations) a vector of length n.
:param fnct: the functional to fit
:param x: the abscissa values
:param y: the ordinate values
:param sd: standard deviation of equations (one or more values used
cyclically)
:param wt: an optional alternate for `sd`
:param fid: the id of the sub-fitter (numerical)
"""
self._fit(fitfunc="linear", fnct=fnct, x=x, y=y, sd=sd, wt=wt, fid=fid)
def _getval(self, valname, fid):
self._checkid(fid)
if not self._fitids[fid]["solved"]:
raise RuntimeError("Not solved yet")
return self._fitids[fid][valname]
def solution(self, fid=0):
"""Return the solution for the fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("sol", fid)
def rank(self, fid=0):
"""Obtain the rank(in SVD sense) of a fit.
The :meth:`constraint` method will show the equations that are
orthogonal to the existing ones, and which will make the solution
possible.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("rank", fid)
def deficiency(self, fid=0):
"""Obtain the missing rank (in SVD sense) of a fit.
The :meth:`constraint` method will show the equations that are
orthogonal to the existing ones, and which will make the solution
possible.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("deficiency", fid)
def chi2(self, fid=0):
"""Obtain the chi squared of a fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("chi2", fid)
def sd(self, fid=0):
"""Obtain the standard deviation per unit of weight of a fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("sd", fid)
def mu(self, fid=0):
"""Obtain the standard deviation per condition equation of a fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("mu", fid)
stddev = mu
def covariance(self, fid=0):
"""Obtain the covariance matrix of a fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("covar", fid)
def error(self, fid=0):
"""Obtain the errors in the unknowns of a fit.
:param fid: the id of the sub-fitter (numerical)
"""
return self._getval("error", fid)
def constraint(self, n=-1, fid=0):
"""Obtain the set of orthogonal equations that make the solution of
the rank deficient normal equations possible.
:param fid: the id of the sub-fitter (numerical)
"""
c = self._getval("constr", fid)
if n < 0 or n > self.deficiency(fid):
return c
else:
raise RuntimeError("Not yet implemented")
def fitted(self, fid=0):
"""Test if enough Levenberg-Marquardt loops have been done.
It returns True if no improvement possible.
:param fid: the id of the sub-fitter (numerical)
"""
self._checkid(fid)
return not (self._fitids[fid]["fit"] > 0
or self._fitids[fid]["fit"] < -0.001)
def _as_array(self, v, dtype=None):
if not hasattr(v, "__len__"):
v = [v]
return NUM.asarray(v, dtype)
python-casacore-2.2.1/casacore/functionals/ 0000775 0000000 0000000 00000000000 13203113151 0020706 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/functionals/__init__.py 0000775 0000000 0000000 00000007023 13203113151 0023024 0 ustar 00root root 0000000 0000000 # __init__.py: Top level .py file for python functionals interface
# Copyright (C) 2006,2007
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: __init__.py,v 1.1 2006/09/29 06:42:55 mmarquar Exp $
"""
Introduction
============
A functional is a function with parameters, defined as *f(p;x)*, where *p* are
the parameters, and *x* the arguments. Methods are available to calculate the
value of a function for a series of argument values for the given set of
parameters, and for the automatic calculation of the derivatives with respect
to the parameters.
The created functionals can be used for fitiing as provided by
:mod:`casacore.fitting`.
A functional has a mask associated with it, to indicate if certain parameters
have to be solved for. See masks for details.
To access the functionals module ``import casacore.functionals``.
Functionals are created in a variety of ways, in general by specifying the
name of the functional, together with some necessary information like e.g.
the order of a polynomial, or the code needed to compile your privately
defined function. Parameters can be set at creation time or later::
>>> from casacore.functionals import functional, gaussian1d
>>> a = gaussian1d() # creates a 1D Gaussian, default arguments
>>> b = functional('gaussian1d') # creates the same one
>>> print a.f(1) # the value at x=1
[0.062500000000000028]
>>> print a(1)
[0.062500000000000028]
>>> print a.fdf([0,0.5]); # value and derivatives
>>> print a([0, 0.5], derivatives=True)
In some cases an order can be specified as well (e.g. for polynomials)::
>>> from casacore.functionals import functional, poly
>>> a = poly(3) # creates a 3rd order polynomial
>>> print a
{'ndim': 1, 'masks': array([ True, True, True, True], dtype=bool),
'params': array([ 1., 1., 1., 1.]), 'npar': 4, 'type': 5, 'order': 3}
An extremely valuable aspect of the Functionals module is the ability to
create a functional from a compiled string specifying an arbitrary function.
For example, let us make our own polynomial ``1 + 2*x + 3*x2`` and evaluate it
at a few abcissa locations::
>>> from casacore.functionals import compiled
>>> a = compiled('p0 + p1*x + p2*x*x', [1,2,3]) # Define
>>> a([0,10,20]) # Evaluate at x=[0,10,20]
[1.0, 321.0, 1241.0]
The functions created can also be used to specify the function to be fitted
in a least squares fit.
"""
from .functional import *
python-casacore-2.2.1/casacore/functionals/functional.py 0000664 0000000 0000000 00000041633 13203113151 0023431 0 ustar 00root root 0000000 0000000 from ._functionals import _functional
import numpy
def copydoc(fromfunc, sep="\n"):
"""
Decorator: Copy the docstring of `fromfunc`
"""
def _decorator(func):
sourcedoc = fromfunc.__doc__
if func.__doc__ is None:
func.__doc__ = sourcedoc
else:
func.__doc__ = sep.join([sourcedoc, func.__doc__])
return func
return _decorator
class functional(_functional):
def __init__(self, name=None, order=-1, params=None, mode=None, dtype=0):
if isinstance(dtype, str):
dtypes = {'real': 0, 'complex': 1}
dtype = dtypes.get(dtype.lower())
if numpy.iscomplexobj(params):
dtype = 1
self._dtype = dtype
progtext = ""
if not isinstance(name, str):
raise TypeError("'name' was not of type string")
if not (isinstance(order, int) or isinstance(order, str)):
raise TypeError("'order' was not of type integer or string")
else:
if isinstance(order, str):
progtext = order
order = -1
# our own functionals server
d = {'type': name, 'order': order, 'progtext': progtext}
if isinstance(mode, dict):
d['mode'] = mode
_functional.__init__(self, d, self._dtype)
if hasattr(params, "__len__"):
params = self._flatten(params)
if len(params) == 0:
pass
elif len(params) == self.npar():
self.set_parameters(params)
else:
raise ValueError("Incorrect number of parameters "
"specified in functional")
def __repr__(self):
return str(self.todict())
def _flatten(self, x):
if (isinstance(x, numpy.ndarray) and x.ndim > 1
and x.ndim == self.ndim()):
return x.flatten()
return x
def ndim(self):
return _functional.ndim(self)
def npar(self):
"""
Return the number of parameters of the functional
:retval: int
"""
return _functional.npar(self)
def __len__(self):
return self.npar()
# def __getitem__(self, i):
# return self.get_parameters()[i]
# def __setitem__(self, i, v):
# return self.set_parameter(i, v)
def set_parameters(self, params):
params = self._flatten(params)
if self._dtype == 0:
return _functional._setparameters(self, params)
else:
return _functional._setparametersc(self, params)
def set_parameter(self, idx, val):
if self._dtype == 0:
return _functional._setpar(self, idx, val)
else:
return _functional._setparc(self, idx, val)
def get_parameters(self):
if self._dtype == 0:
return _functional._parameters(self)
else:
return _functional._parametersc(self)
def f(self, x):
"""Calculate the value of the functional for the specified arguments
(taking any specified mask into account).
:param x: the value(s) to evaluate at
"""
x = self._flatten(x)
if self._dtype == 0:
return numpy.array(_functional._f(self, x))
else:
return numpy.array(_functional._fc(self, x))
def __call__(self, x, derivatives=False):
if derivatives:
return numpy.array(self.fdf(x))
else:
return numpy.array(self.f(x))
def fdf(self, x):
"""Calculate the value of the functional for the specified arguments,
and the derivatives with respect to the parameters (taking any
specified mask into account).
:param x: the value(s) to evaluate at
"""
x = self._flatten(x)
n = 1
if hasattr(x, "__len__"):
n = len(x)
if self._dtype == 0:
retval = _functional._fdf(self, x)
else:
retval = _functional._fdfc(self, x)
if len(retval) == n:
return numpy.array(retval)
return numpy.array(retval).reshape(self.npar() + 1,
n // self.ndim()).transpose()
def add(self, other):
if not isinstance(other, functional):
raise TypeError("'other' is not a functional")
if self._dtype != other._dtype:
raise TypeError("'other' is not of the same value type")
if self._dtype == 0:
_functional._add(self, other)
else:
_functional._addc(self, other)
def set_mask(self, i, msk):
_functional._setmask(self, i, msk)
def set_masks(self, msk):
_functional._setmasks(self, msk)
def get_masks(self):
return _functional._masks(self)
def todict(self):
return _functional.todict(self)
class gaussian1d(functional):
"""Create a 1-dimensional Gaussian with the specified height, width and
center.
:param params: the [height, center, width] as a list
"""
def __init__(self, params=None, dtype=0):
functional.__init__(self, name="gaussian1d", params=params,
dtype=dtype)
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> g = gaussian1d([1, 2, 3])
>>> print g.npar()
3
>>> print len(g)
3
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> a = gaussian1d()
>>> print(a.f(0.0))
[ 1.]
>>> print(a(0.0)) #equivalent
[ 1.]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> g = gaussian1d()
>>> print(g.fdf(0.0))
[[ 1., 1., 0., 0.]]
>>> print(g(0.0, derivatives=True)) #equivalent
[[ 1., 1., 0., 0.]]
"""
return functional.fdf(self, x)
class gaussian2d(functional):
"""
Create a two-dimensional gaussian.
:param params: list [amplitude, centers, major width, ratio, angle] of
Gaussian default is [1, 0, 0, 1, 1, 0]
:param dtype: The data type. One of 'real' or 0, or 'complex' or 1
"""
def __init__(self, params=None, dtype=0):
if params is None:
params = [1, 0, 0, 1, 1, 0]
functional.__init__(self, name="gaussian2d",
params=params,
dtype=dtype)
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> g = gaussian2d([1, 2, 3][4, 5, 6])
>>> print g.npar()
6
>>> print len(g)
6
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> a = gaussian2d()
>>> print(a.f(0.0))
[]
>>> print(a(0.0)) #equivalent
[]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> a = gaussian2d()
>>> print(a.fdf(0))
[]
>>> print(g(0.0, derivatives=True)) #equivalent
[]
"""
return functional.fdf(self, x)
class poly(functional):
"""
Create a polynomial of specified degree. The default parameters are all 1.
(Note that using the generic functional function the parameters are all
set to 0).
:param order: the order of the polynomial (number of parameters -1)
:param params: the values of the parameters as a list.
:param dtype: the optional data type. Default is float, but will be
auto-detected from `params`. Can be set to 'complex'.
"""
def __init__(self, order, params=None, dtype=0):
functional.__init__(self, name="poly",
order=order,
params=params,
dtype=dtype)
if params is None:
self.set_parameters([v + 1. for v in self.get_parameters()])
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> p = poly(5)
>>> print p.npar()
6
>>> print len(p)
6
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> p = poly(5)
>>> print(p.f(0.0))
[ 1.]
>>> print(p(0.0)) # equivalent
[ 1.]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> p = poly(5)
>>> print(p.fdf(0.0))
[[ 1., 1., 0., 0., 0., 0., 0.]]
>>>print(p(0.0, derivatives=True)) # equivalent
[[ 1., 1., 0., 0., 0., 0., 0.]]
"""
return functional.fdf(self, x)
class oddpoly(functional):
"""Create an odd polynomial of specified degree.
:param order: the order of the polynomial
:param params: the values of the parameters as a list.
:param dtype: the optional data type. Default is float, but will be
auto-detected from `params`. Can be set to 'complex'.
"""
def __init__(self, order, params=None, dtype=0):
functional.__init__(self, name="oddpoly",
order=order,
params=params,
dtype=dtype)
if params is None:
self.set_parameters([v + 1. for v in self.get_parameters()])
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> p = oddpoly(3)
>>> print p.npar()
2
>>> print len(p)
2
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> p = oddpoly(3)
>>> print(p.f(0.0))
[ 0.]
>>> print(p(0.0)) # equivalent
[ 0.]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> p = oddpoly(3)
>>> print(p.fdf(0.0))
[[ 0., 0., 0.]]
>>> print(p(0.0, derivatives=True)) # equivalent
[[ 0., 0., 0.]]
"""
return functional.fdf(self, x)
class evenpoly(functional):
"""Create an even polynomial of specified degree.
:param order: the order of the polynomial
:param params: the values of the parameters as a list.
:param dtype: the optional data type. Default is float, but will be
auto-detected from `params`. Can be set to 'complex'.
"""
def __init__(self, order, params=None, dtype=0):
functional.__init__(self, name="evenpoly",
order=order,
params=params,
dtype=dtype)
if params is None:
self.set_parameters([v + 1. for v in self.get_parameters()])
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> p = evenpoly(2)
>>> print p.npar()
2
>>> print len(p)
2
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> p = evenpoly(2)
>>> print(p.f(0.0))
[ 1.]
>>> print(p(0.0)) # equivalent
[ 1.]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> p = oddpoly(3)
>>> print(p.fdf(0.0))
[[ 1., 1., 0.]]
>>>print(p(0.0, derivatives=True)) # equivalent
[[ 1., 1., 0.]]
"""
return functional.fdf(self, x)
class chebyshev(functional):
def __init__(self, order, params=None,
xmin=-1., xmax=1., ooimode='constant',
dtype=0):
modes = "constant zeroth extrapolate cyclic edge".split()
if ooimode not in modes:
raise ValueError("Unrecognized ooimode")
mode = {'interval': [float(xmin), float(xmax)], 'intervalMode': ooimode,
'default': float(0.0)}
functional.__init__(self, name="chebyshev",
order=order,
params=params,
mode=mode,
dtype=dtype)
if params is None:
self.set_parameters([v + 1. for v in self.get_parameters()])
@copydoc(functional.npar)
def npar(self):
"""
Equivalent::
>>> ch = chebyshev(2)
>>> print ch.npar()
4
>>> print len(p)
4
"""
return functional.npar(self)
@copydoc(functional.f)
def f(self, x):
"""
Example::
>>> ch = chebyshev(2)
>>> print(ch.f(0.0))
[ 0.]
>>> print(ch(0.0)) # equivalent
[ 0.]
"""
return functional.f(self, x)
@copydoc(functional.fdf)
def fdf(self, x):
"""
Example::
>>> ch = chebyshev(2)
>>> print(ch.fdf(0.0))
[[ 0., 1., 0., -1.]]
>>>print(ch(0.0, derivatives=True)) # equivalent
[[ 0., 1., 0., -1.]]
"""
return functional.fdf(self, x)
class compound(functional):
def __init__(self, dtype=0):
"""Create a compound function.
This class takes a arbitary number of functions and
generates a new single function object.
Example::
>>> d = poly(2)
>>> gauss1d = gaussian1d([1, 0, 1])
>>> sum = compound()
>>> sum.add(d)
>>> sum.add(gauss1d)
>>> print(sum(2))
[ 7.00001526]
"""
functional.__init__(self, name="compound", dtype=dtype)
class combi(functional):
def __init__(self, dtype=0):
"""Form a linear combinations of functions object.
Example::
>>> const = poly(0)
>>> linear = poly(1)
>>> square = poly(2)
>>> c = combi()
>>> c.add(const)
>>> c.add(linear)
>>> c.add(square)
>>> print(c(0))
[ 3.]
"""
functional.__init__(self, name="combi", dtype=dtype)
class compiled(functional):
"""Create a function based on the programable string. The string should
be a single expression, which can use the standard operators and
functions and parentheses, having a single value as a result. The
parameters of the function can be addressed with the *p* variable. This
variable can be indexed in two ways. The first way is using the standard
algebraic way, where the parameters are: ``p (or p0), p1, p2, ...`` . The
second way is by indexing, where the parameters are addressed as: p[0],
p[1], ... . The arguments are accessed in the same way, but using the
variable name x. The compilation determines the number of dimensions and
parameters of the produced function.
Operators are the standard operators (including comparisons, which
produce a zero or one result; and conditional expression).
In addition to the standard expected functions, there is an atan with
either one or two arguments (although atan2 exists as well), and pi and
ee with no or one argument. The functional created behaves as all other
functionals, and hence can be used in combinations.
Examples::
>>> from casacore.functionals import compiled
>>> import math
>>> a = compiled('sin(pi(0.5) ) +pi'); # an example
>>> print a(0)
array([ 4.1415926535897931])
>>> b = compiled('p*exp(-(x/p[2])^2)')
>>> print b.get_parameters()
[0.0, 0.0]
>>> b.set_parameters([10, 1]) # change to height 10 and halfwidth 1
>>> print b([-1,-0.5,0,.5,1])
array([ 3.6787944117144233,
7.788007830714049,
10.0,
7.788007830714049,
3.6787944117144233])
# the next one is sync(x), catering for x=0
# using the fact that comparisons deliver values. Note
# the extensive calculation to make sure no divison by 0
>>> synca = compiled('( (x==0) * 1)+( (x!=0) * sin(x+(x==0)*1)/(x+(x==0)*1) )')
>>> print synca([-1,0,1])
[0.841471, 1., 0.841471]
>>> print math.sin(1)/1
0.841471
# using conditional expressions:
print compiled('x==0 ? 1 : sin(x)/x')([-1,0,1])
[0.841471, 1.0, 0.841471]
"""
def __init__(self, code="", params=None, dtype=0):
functional.__init__(self, name="compiled", order=code,
params=params, dtype=dtype)
python-casacore-2.2.1/casacore/images/ 0000775 0000000 0000000 00000000000 13203113151 0017626 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/images/__init__.py 0000664 0000000 0000000 00000004377 13203113151 0021752 0 ustar 00root root 0000000 0000000 # __init__.py: Python image functions
# Copyright (C) 2008
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id$
"""Python interface to the Casacore images module.
A `casacore image <../../casacore/doc/html/group__Images__module.html>`_
represents an astronomical image of arbitrary dimensionality.
Several image formats are recognized:
`casacore paged image <../../casacore/doc/html/classcasa_1_1PagedImage.html>`_
is the native casacore image format stored in a casacore table.
`HDF5 `_
is the HDF5 format often used in the earth science community.
`FITS `_
is the well-known astronomical FITS format
`miriad `_
is the format used by the radio-astronomical MIRIAD package.
The following functionality exists:
- get and put data (slices)
- get or put a mask
- get meta data like coordinates and history
- get, put, or search optional image attributes (as used for LOFAR)
- get statistics
- form a subimage
- form an image expression which is treated as an ordinary image
- regrid the image
- write the image to a FITS file
"""
# Make image interface available.
from .image import image
python-casacore-2.2.1/casacore/images/coordinates.py 0000664 0000000 0000000 00000026371 13203113151 0022523 0 ustar 00root root 0000000 0000000 # coordinates.py: Python coordinate system wrapper
# Copyright (C) 2008
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id$
from casacore import six
class coordinatesystem(object):
"""
A thin wrapper for casacore coordinate systems. It dissects the
coordinatesystem record returned from casacore images.
This only handles one instance of each coordinate type.
The possible types are ''direction'', ''spectral'', ''stokes'',
''linear'' and ''tabular''.
The first, second, trird and fourth axis are respectively,
'Right Ascension','Declination','Stokes' and 'Frequency'.
To make a coordinate object, these things should be taken care
of.Like to make a spectral coordinate, a 4D image should be used
as the 4th axis is 'Frequency' and so on.
It uses reference semantics for the individual coordinates,
e.g. the following will work::
cs = im.coordinates()
cs["direction"].set_referencepixel([0.0,0.0])
# or equivalent
cs.get_coordinates("direction").set_referencepixel([0.0,0.0])
"""
def __init__(self, rec):
self._csys = rec
self._names = []
self._get_coordinatenames()
def __str__(self):
out = ""
for coord in self:
out += str(coord)
return out
def dict(self):
return self._csys
def summary(self):
six.print_(str(self))
def _get_coordinatenames(self):
"""Create ordered list of coordinate names
"""
validnames = ("direction", "spectral", "linear", "stokes", "tabular")
self._names = [""] * len(validnames)
n = 0
for key in self._csys.keys():
for name in validnames:
if key.startswith(name):
idx = int(key[len(name):])
self._names[idx] = name
n += 1
# reverse as we are c order in python
self._names = self._names[:n][::-1]
if len(self._names) == 0:
raise LookupError("Coordinate record doesn't contain valid coordinates")
def get_names(self):
"""Get the coordinate names
"""
return self._names
def __getitem__(self, name):
# reverse index back to fortran order as the record is using this
i = self._names[::-1].index(name)
return eval("%scoordinate(self._csys['%s'])" % (name, name + str(i)))
# alias
get_coordinate = __getitem__
def __setitem__(self, name, val):
# reverse index back to fortran order as the record is using this
i = self._names[::-1].index(name)
assert isinstance(val, eval("%scoordinate" % name))
self._csys[name + str(i)] = val._coord
# alias
set_coordinate = __setitem__
def __iter__(self):
for name in self._names:
yield self.get_coordinate(name)
def get_obsdate(self):
return self._csys.get("obsdate", None)
def get_observer(self):
return self._csys.get("observer", None)
def get_telescope(self):
return self._csys.get("telescope", None)
def get_referencepixel(self):
return [coord.get_referencepixel() for coord in self]
def set_referencepixel(self, values):
for i, coord in enumerate(self):
coord.set_referencepixel(values[i])
def get_referencevalue(self):
return [coord.get_referencevalue() for coord in self]
def set_referencevalue(self, values):
for i, coord in enumerate(self):
coord.set_referencevalue(values[i])
def get_increment(self):
return [coord.get_increment() for coord in self]
def set_increment(self, values):
for i, coord in enumerate(self):
coord.set_increment(values[i])
def get_unit(self):
return [coord.get_unit() for coord in self]
def get_axes(self):
return [coord.get_axes() for coord in self]
class coordinate(object):
"""Overwrite as neccessary
"""
def __init__(self, rec):
self._coord = rec
self._template = " %-16s: %s\n"
def __str__(self):
lname = self.__class__.__name__.capitalize()
out = "%s Coordinate:\n" % lname[:-10]
out += self._template % ("Reference Pixel",
str(self.get_referencepixel()))
out += self._template % ("Reference Value",
str(self.get_referencevalue()) \
+ " " + str(self.get_unit()))
out += self._template % ("Increment",
str(self.get_increment()) \
+ " " + str(self.get_unit()))
return out
def dict(self):
"""Get the coordinate info as a dict"""
return self._coord
def get_axis_size(self, axis=0):
"""Get the length of the given axis in this coordinate
-1 is returned if unknown.
"""
try:
return self._coord["_axes_sizes"][axis]
except:
return -1
def get_image_axis(self, axis=0):
"""Get the image axis number of the given axis in this coordinate
-1 is returned if unknown.
"""
try:
return self._coord["_image_axes"][axis]
except:
return -1
# ALL list/array values have to be reversed as the coordsys dict holds
# everything in fortran order.
def get_referencepixel(self):
"""Get the reference pixel of the given axis in this coordinate."""
return self._coord.get("crpix", [])[::-1]
def set_referencepixel(self, pix):
"""Set the reference pixel of the given axis in this coordinate."""
assert len(pix) == len(self._coord["crpix"])
self._coord["crpix"] = pix[::-1]
def get_referencevalue(self):
"""Get the reference value of the given axis in this coordinate."""
return self._coord.get("crval", [])[::-1]
def set_referencevalue(self, val):
"""Set the reference pixel of the given axis in this coordinate."""
assert len(val) == len(self._coord["crval"])
self._coord["crval"] = val[::-1]
def get_increment(self):
"""Get the increment of the given axis in this coordinate."""
return self._coord.get("cdelt", [])[::-1]
def set_increment(self, inc):
"""Set the increment of the given axis in this coordinate."""
self._coord["cdelt"] = inc[::-1]
def get_unit(self):
"""Get the unit of the given axis in this coordinate."""
return self._coord.get("units", [])[::-1]
def get_axes(self):
"""Get the axes in this coordinate."""
return self._coord.get("axes", [])[::-1]
class directioncoordinate(coordinate):
def __init__(self, rec):
coordinate.__init__(self, rec)
def __str__(self):
out = coordinate.__str__(self)
out += self._template % ("Frame", str(self.get_frame()))
out += self._template % ("Projection", str(self.get_projection()))
return out
def get_projection(self):
"""Get the projection of the given axis in this coordinate."""
return self._coord.get("projection", None)
def set_projection(self, val):
"""Set the projection of the given axis in this coordinate.
The known projections are SIN, ZEA, TAN, NCP, AIT, ZEA
"""
knownproj = ["SIN", "ZEA", "TAN", "NCP", "AIT", "ZEA"] # etc
assert val.upper() in knownproj
self._coord["projection"] = val.upper()
def get_frame(self):
return self._coord.get("system", None)
def set_frame(self, val):
# maybe uses measures here
# dm = measures();knonwframes = dm.list_codes(dm.direction())["normal"]
knownframes = ["GALACTIC", "J2000", "B1950", "SUPERGAL"] # etc
assert val.upper() in knownframes
self._coord["system"] = val.upper()
class spectralcoordinate(coordinate):
def __init__(self, rec):
coordinate.__init__(self, rec)
def __str__(self):
out = coordinate.__str__(self)
out += self._template % ("Frame", str(self.get_frame()))
out += self._template % ("Rest Frequency",
str(self.get_restfrequency()) + " Hz")
return out
def get_unit(self):
return self._coord.get("unit", None)
def get_referencepixel(self):
return self._coord["wcs"].get("crpix", None)
def set_referencepixel(self, pix):
self._coord["wcs"]["crpix"] = pix
def get_referencevalue(self):
return self._coord["wcs"].get("crval", None)
def set_referencevalue(self, val):
self._coord["wcs"]["crval"] = val
def get_increment(self):
return self._coord["wcs"].get("cdelt", None)
def set_increment(self, inc):
self._coord["wcs"]["cdelt"] = inc
def get_axes(self):
return self._coord.get("name", None)
def get_restfrequency(self):
return self._coord.get("restfreq", None)
def set_restfrequency(self, val):
self._coord["restfreq"] = val
def get_frame(self):
return self._coord.get("system", None)
def set_frame(self, val):
# maybe uses measures here
# dm = measures();knonwframes = dm.list_codes(dm.frequency())["normal"]
knownframes = ["BARY", "LSRK", "TOPO"]
assert val.upper() in knownframes
self._coord["system"] = val.upper()
def get_conversion(self):
return self._coord.get("conversion", None)
def set_conversion(self, key, val):
assert key in self._coord
self._coord["conversion"][key] = val
class linearcoordinate(coordinate):
def __init__(self, rec):
coordinate.__init__(self, rec)
class stokescoordinate(coordinate):
def __init__(self, rec):
coordinate.__init__(self, rec)
def get_stokes(self):
return self._coord["stokes"]
class tabularcoordinate(coordinate):
def __init__(self, rec):
coordinate.__init__(self, rec)
def get_pixelvalues(self):
return self._coord["pixelvalues"]
def set_pixelvalues(self, val):
assert len(val) == len(self._coord["pixelvalues"])
self._coord["pixelvalues"] = val
def get_worldvalues(self):
return self._coord["worldvalues"]
def set_worldvalues(self, val):
assert len(val) == len(self._coord["worldvalues"])
self._coord["worldvalues"] = val
python-casacore-2.2.1/casacore/images/image.py 0000664 0000000 0000000 00000064323 13203113151 0021272 0 ustar 00root root 0000000 0000000 # image.py: Python image functions
# Copyright (C) 2008
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id$
# Make interface to class ImageProxy available.
from ._images import Image
import numpy
try:
import numpy.ma as nma
except ImportError:
import numpy.core.ma as nma
from casacore.images.coordinates import coordinatesystem
from casacore import six
class image(Image):
"""The Python interface to casacore images.
An image can be constructed in a variety of ways:
- Opening an existing image. The image format is determined automatically
and can be
`casacore `_,
`HDF5 `_,
`FITS `_,
or `MIRIAD `_.
FITS and MIRIAD always have data type float, but casacore and HDF5 images
can have data type float, double, complex, or dcomplex.
- Open an image expression by giving a `LEL expression
<../../casacore/doc/notes/223.html>`_ string.
Note that in an expression `$im` can be used similar to TaQL commands
(see function :func:`tables.taql`).
- Create a new temporary image from a shape or a numpy array.
- Virtually concatenate a number of images along a given axis. This can
be used to form a spectral line image cube from separate channel images.
The following arguments can be given:
`imagename`
| It can be given in several forms:
| If it is a tuple or list, the image is opened as the virtual
concatenation of the given image names or objects.
| Otherwise it should be a string giving the image name (or expression).
If argument `values` or `shape` is given, a new image with that name
is created using the values or shape. If the name is empty, a temporary
image is created which can be written later using :func:`saveas`.
| Otherwise it is tried to open an existing image with that name.
| If the open fails, it is opened as a LEL image expression.
`axis`
The axis number along which images have to be concatenated.
`maskname`
| The name of the mask to be used when opening an existing image. If not
given, the default image mask is used.
| If an image is created, a mask with this name is created.
`images`
Possible images objects to be used for $n arguments in an expression
string.
'values`
If given, the image will be created and these values will be stored in it.
It should be a numpy array or masked array. If it is a masked array,
its mask acts as the `mask` argument described below.
The data type of the image is derived from the array's data type.
`coordsys`
The coordinate system to be used when creating an image.
If not given, an appropriate default coordnate system will be used.
`overwrite`
If False, an exception is raised if the new image file already exists.
Default is True.
`ashdf5`
If True, the image is created in HDF5 format, otherwise in casacore
format. Default is casacore format.
`mask`
| An optional mask to be stored in the image when creating the image.
If a mask is given, but no maskname is given, the mask will get the
name `mask0`.
| Note that the mask can also be given in argument `values` (see above).
`shape`
If given, the image will be created. If `values` is also given, its
shape should match. If `values` is not given, an image with data type
double will be created.
`tileshape`
Advanced users can give the tile shape to be used. See the
:mod:`tables` module for more information about Tiled Storage Managers.
For example::
im = image('3c343.fits') # open existing fits image
im = image('a.img1 - a.img2') # open as expression
im = image(shape=(256,256)) # create temp image
im = image('a', shape=(256,256)) # create image a
"""
def __init__(self, imagename, axis=0, maskname="", images=(), values=None,
coordsys=None, overwrite=True, ashdf5=False, mask=(),
shape=None, tileshape=()):
coord = {}
if coordsys is not None:
coord = coordsys.dict()
if isinstance(imagename, Image):
# Create from the value returned by subimage, etc.
Image.__init__(self, imagename)
else:
opened = False
if isinstance(imagename, tuple) or isinstance(imagename, list):
if len(imagename) == 0:
raise ValueError('No images given in list or tuple')
if isinstance(imagename[0], str):
# Concatenate from image names
Image.__init__(self, imagename, axis)
opened = True
elif isinstance(imagename[0], image):
# Concatenate from image objects
Image.__init__(self, imagename, axis, 0, 0)
opened = True
if not opened:
if not isinstance(imagename, str):
raise ValueError("first argument must be name or" +
" sequence of images or names")
if shape is None:
if values is None:
# Open an image from name or expression
# Copy the tables argument and make sure it is a list
imgs = []
for img in images:
imgs += [img]
try:
# Substitute possible $ arguments
import casacore.util as cu
imagename = cu.substitute(imagename,
[(image, '', imgs)],
locals=cu.getlocals(3))
except:
six.print_("Probably could not import casacore.util")
pass
Image.__init__(self, imagename, maskname, imgs)
else:
# Create an image from an array
# The values can be a masked array
# use the mask if no explicit mask is given
if isinstance(values, nma.MaskedArray):
if len(mask) == 0:
mask = nma.getmaskarray(values)
values = values.data
if len(mask) > 0:
mask = -mask # casa and numpy have opposite flags
Image.__init__(self, values, mask, coord,
imagename, overwrite, ashdf5,
maskname, tileshape)
else:
# Create an image from a shape (values gives the data type)
# default type is float.
if values is None:
values = numpy.array([0], dtype='float32')[0]
Image.__init__(self, shape, values, coord,
imagename, overwrite, ashdf5,
maskname, tileshape, 0)
def __str__(self):
"""Get image name."""
return self.name(strippath=True)
def __len__(self):
"""Get nr of pixels in the image."""
return self._size()
def ispersistent(self):
"""Test if the image is persistent, i.e. stored on disk."""
return self._ispersistent()
def name(self, strippath=False):
"""Get image name."""
return self._name(strippath)
def shape(self):
"""Get image shape."""
return self._shape()
def ndim(self):
"""Get dimensionality of the image."""
return self._ndim()
def size(self):
"""Get nr of pixels in the image."""
return self._size()
def datatype(self):
"""Get data type of the image."""
return self._datatype()
def imagetype(self):
"""Get image type of the image (PagedImage, HDF5Image, etc.)."""
return self._imagetype()
def attrgroupnames(self):
"""Get the names of all attribute groups."""
return self._attrgroupnames()
def attrcreategroup(self, groupname):
"""Create a new attribute group."""
self._attrcreategroup(groupname)
def attrnames(self, groupname):
"""Get the names of all attributes in this group."""
return self._attrnames(groupname)
def attrnrows(self, groupname):
"""Get the number of rows in this attribute group."""
return self._attrnrows(groupname)
def attrget(self, groupname, attrname, rownr):
"""Get the value of an attribute in the given row in a group."""
return self._attrget(groupname, attrname, rownr)
def attrgetcol(self, groupname, attrname):
"""Get the value of an attribute for all rows in a group."""
values = []
for rownr in range(self.attrnrows(groupname)):
values.append(self.attrget(groupname, attrname, rownr))
return values
def attrfindrows(self, groupname, attrname, value):
"""Get the row numbers of all rows where the attribute matches the given value."""
values = self.attrgetcol(groupname, attrname)
return [i for i in range(len(values)) if values[i] == value]
def attrgetrow(self, groupname, key, value=None):
"""Get the values of all attributes of a row in a group.
If the key is an integer, the key is the row number for which
the attribute values have to be returned.
Otherwise the key has to be a string and it defines the name of an
attribute. The attribute values of the row for which the key matches
the given value is returned.
It can only be used for unique attribute keys. An IndexError exception
is raised if no or multiple matches are found.
"""
if not isinstance(key, str):
return self._attrgetrow(groupname, key)
# The key is an attribute name whose value has to be found.
rownrs = self.attrfindrows(groupname, key, value)
if len(rownrs) == 0:
raise IndexError("Image attribute " + key + " in group " +
groupname + " has no matches for value " +
str(value))
if len(rownrs) > 1:
raise IndexError("Image attribute " + key + " in group " +
groupname + " has multiple matches for value " +
str(value))
return self._attrgetrow(groupname, rownrs[0])
def attrgetunit(self, groupname, attrname):
"""Get the unit(s) of an attribute in a group."""
return self._attrgetunit(groupname, attrname)
def attrgetmeas(self, groupname, attrname):
"""Get the measinfo (type, frame) of an attribute in a group."""
return self._attrgetmeas(groupname, attrname)
def attrput(self, groupname, attrname, rownr, value, unit=[], meas=[]):
"""Put the value and optionally unit and measinfo
of an attribute in a row in a group."""
return self._attrput(groupname, attrname, rownr, value, unit, meas)
def getdata(self, blc=(), trc=(), inc=()):
"""Get image data.
Using the arguments blc (bottom left corner), trc (top right corner),
and inc (stride) it is possible to get a data slice.
The data is returned as a numpy array. Its dimensionality is the same
as the dimensionality of the image, even if an axis has length 1.
"""
return self._getdata(self._adjustBlc(blc),
self._adjustTrc(trc),
self._adjustInc(inc))
# Negate the mask; in numpy True means invalid.
def getmask(self, blc=(), trc=(), inc=()):
"""Get image mask.
Using the arguments blc (bottom left corner), trc (top right corner),
and inc (stride) it is possible to get a mask slice. Not all axes
need to be specified. Missing values default to begin, end, and 1.
The mask is returned as a numpy array. Its dimensionality is the same
as the dimensionality of the image, even if an axis has length 1.
Note that the casacore images use the convention that a mask value
True means good and False means bad. However, numpy uses the opposite.
Therefore the mask will be negated, so it can be used directly in
numpy operations.
If the image has no mask, an array will be returned with all values
set to False.
"""
return numpy.logical_not(self._getmask(self._adjustBlc(blc),
self._adjustTrc(trc),
self._adjustInc(inc)))
# Get data and mask
def get(self, blc=(), trc=(), inc=()):
"""Get image data and mask.
Get the image data and mask (see ::func:`getdata` and :func:`getmask`)
as a numpy masked array.
"""
return nma.masked_array(self.getdata(blc, trc, inc),
self.getmask(blc, trc, inc))
def putdata(self, value, blc=(), trc=(), inc=()):
"""Put image data.
Using the arguments blc (bottom left corner), trc (top right corner),
and inc (stride) it is possible to put a data slice. Not all axes
need to be specified. Missing values default to begin, end, and 1.
The data should be a numpy array. Its dimensionality must be the same
as the dimensionality of the image.
"""
return self._putdata(value, self._adjustBlc(blc),
self._adjustInc(inc))
def putmask(self, value, blc=(), trc=(), inc=()):
"""Put image mask.
Using the arguments blc (bottom left corner), trc (top right corner),
and inc (stride) it is possible to put a data slice. Not all axes
need to be specified. Missing values default to begin, end, and 1.
The data should be a numpy array. Its dimensionality must be the same
as the dimensionality of the image.
Note that the casacore images use the convention that a mask value
True means good and False means bad. However, numpy uses the opposite.
Therefore the mask will be negated, so a numoy masked can be given
directly.
The mask is not written if the image has no mask and if it the entire
mask is False. In that case the mask most likely comes from a getmask
operation on an image without a mask.
"""
# casa and numpy have opposite flags
return self._putmask(~value, self._adjustBlc(blc),
self._adjustInc(inc))
def put(self, value, blc=(), trc=(), inc=()):
"""Put image data and mask.
Put the image data and optionally the mask (see ::func:`getdata`
and :func:`getmask`).
If the `value` argument is a numpy masked array, but data and mask will
bw written. If it is a normal numpy array, only the data will be
written.
"""
if isinstance(value, nma.MaskedArray):
self.putdata(value.data, blc, trc, inc)
self.putmask(nma.getmaskarray(value), blc, trc, inc)
else:
self.putdata(value, blc, trc, inc)
def haslock(self, write=False):
"""Test if the image holds a read or write lock.
| See `func:`tables.table.haslock` for more information.
| Locks are only used for images in casacore format. For other formats
(un)locking is a no-op, so this method always returns True.
"""
return self._haslock(write)
def lock(self, write=False, nattempts=0):
"""Acquire a read or write lock on the image.
| See `func:`tables.table.haslock` for more information.
| Locks are only used for images in casacore format. For other formats
(un)locking is a no-op, so this method always returns True.
Only advanced users should use locking. In normal operations
explicit locking and unlocking is not necessary.
"""
return self._lock(write, nattempts)
def unlock(self):
"""Release a lock on the image.
| See `func:`tables.table.haslock` for more information.
| Locks are only used for images in casacore format. For other formats
(un)locking is a no-op, so this method always returns True.
"""
return self._unlock()
def subimage(self, blc=(), trc=(), inc=(), dropdegenerate=True):
"""Form a subimage.
An image object containing a subset of an image is returned.
The arguments blc (bottom left corner), trc (top right corner),
and inc (stride) define the subset. Not all axes need to be specified.
Missing values default to begin, end, and 1.
By default axes with length 1 are left out.
A subimage is a so-called virtual image. It is not stored, but only
references the original image. It can be made persistent using the
:func:`saveas` method.
"""
return image(self._subimage(self._adjustBlc(blc),
self._adjustTrc(trc),
self._adjustInc(inc),
dropdegenerate))
def coordinates(self):
"""Get the :class:`coordinatesystem` of the image."""
return coordinatesystem(self._coordinates())
def toworld(self, pixel):
"""Convert the pixel coordinates of an image value to world coordinates.
The coordinates must be given with the slowest varying axis first.
Thus normally like frequency(,polarisation axis),Dec,Ra.
"""
return self._toworld(pixel, True)
def topixel(self, world):
"""Convert the world coordinates of an image value to pixel coordinates.
The coordinates must be given with the slowest varying axis first.
Thus normally like frequency(,polarisation axis),Dec,Ra.
"""
return self._topixel(world, True)
def imageinfo(self):
"""Get the standard image info."""
return self._imageinfo()
def miscinfo(self):
"""Get the auxiliary image info."""
return self._miscinfo()
def unit(self):
"""Get the pixel unit of the image."""
return self._unit()
def history(self):
"""Get the image processing history."""
return self._history()
def info(self):
"""Get coordinates, image info, and unit"."""
return {'coordinates': self._coordinates(),
'imageinfo': self._imageinfo(),
'miscinfo': self._miscinfo(),
'unit': self._unit()
}
def tofits(self, filename, overwrite=True, velocity=True,
optical=True, bitpix=-32, minpix=1, maxpix=-1):
"""Write the image to a file in FITS format.
`filename`
FITS file name
`overwrite`
If False, an exception is raised if the new image file already exists.
Default is True.
`velocity`
By default a velocity primary spectral axis is written if possible.
`optical`
If writing a velocity, use the optical definition
(otherwise use radio).
`bitpix`
can be set to -32 (float) or 16 (short) only. When `bitpix` is
16 it will write BSCALE and BZERO into the FITS file. If minPix
`minpix` and `maxpix`
are used to determine BSCALE and BZERO if `bitpix=16`.
If `minpix` is greater than `maxpix` (which is the default),
the minimum and maximum pixel values will be determined from the ddta.
Oherwise the supplied values will be used and pixels outside that
range will be clipped to the minimum and maximum pixel values.
Note that this truncation does not occur for `bitpix=-32`.
"""
return self._tofits(filename, overwrite, velocity, optical,
bitpix, minpix, maxpix)
def saveas(self, filename, overwrite=True, hdf5=False,
copymask=True, newmaskname="", newtileshape=()):
"""Write the image to disk.
Note that the created disk file is a snapshot, so it is not updated
for possible later changes in the image object.
`overwrite`
If False, an exception is raised if the new image file already exists.
Default is True.
`ashdf5`
If True, the image is created in HDF5 format, otherwise in casacore
format. Default is casacore format.
`copymask`
By default the mask is written as well if the image has a mask.
'newmaskname`
If the mask is written, the name is the same the original or
`mask0` if the original mask has no name. Using this argument a
different mask name can be given.
`tileshape`
Advanced users can give a new tile shape. See the :mod:`tables`
module for more information about Tiled Storage Managers.
"""
self._saveas(filename, overwrite, hdf5,
copymask, newmaskname,
newtileshape)
def statistics(self, axes=(), minmaxvalues=(), exclude=False, robust=True):
"""Calculate statistics for the image.
Statistics are returned in a dict for the given axes.
E.g. if axes [0,1] is given in a 3-dim image, the statistics are
calculated for each plane along the 3rd axis. By default statistics
are calculated for the entire image.
`minmaxvalues` can be given to include or exclude pixels
with values in the given range. If only one value is given,
min=-abs(val) and max=abs(val).
By default robust statistics (Median, MedAbsDevMed, and Quartile) are
calculated too.
"""
return self._statistics(self._adaptAxes(axes), "",
minmaxvalues, exclude, robust)
def regrid(self, axes, coordsys, outname="", overwrite=True,
outshape=(), interpolation="linear",
decimate=10, replicate=False,
refchange=True, forceregrid=False):
"""Regrid the image to a new image object.
Regrid the image on the given axes to the given coordinate system.
The output is stored in the given file; it no file name is given a
temporary image is made.
If the output shape is empty, the old shape is used.
`replicate=True` means replication rather than regridding.
"""
return image(self._regrid(self._adaptAxes(axes),
outname, overwrite,
outshape, coordsys.dict(),
interpolation, decimate, replicate,
refchange, forceregrid))
def view(self, tempname='/tmp/tempimage'):
"""Display the image using casaviewer.
If the image is not persistent, a copy will be made that the user
has to delete once viewing has finished. The name of the copy can be
given in argument `tempname`. Default is '/tmp/tempimage'.
"""
import os
# Test if casaviewer can be found.
# On OS-X 'which' always returns 0, so use test on top of it.
if os.system('test -x `which casaviewer` > /dev/null 2>&1') == 0:
six.print_("Starting casaviewer in the background ...")
self.unlock()
if self.ispersistent():
os.system('casaviewer ' + self.name() + ' &')
elif len(tempname) > 0:
six.print_(" making a persistent copy in " + tempname)
six.print_(" which should be deleted after the viewer has ended")
self.saveas(tempname)
os.system('casaviewer ' + tempname + ' &')
else:
six.print_("Cannot view because the image is in memory only.")
six.print_("You can browse a persistent copy of the image like:")
six.print_(" t.view('/tmp/tempimage')")
else:
six.print_("casaviewer cannot be found")
def _adaptAxes(self, axes):
# If axes is a single integer value, turn it into a list.
if isinstance(axes, int):
axes = [axes]
# ImageProxy expects Fortran-numbered axes.
# So reverse the axes.
n = self.ndim() - 1
axout = []
for i in range(len(axes), 0, -1):
axout += [n - axes[i - 1]]
return axout
def _adjust(self, val, defval):
retval = defval
if isinstance(val, tuple) or isinstance(val, list):
retval[0:len(val)] = val
else:
retval[0] = val
return retval
# Append blc with 0 if shorter than shape.
def _adjustBlc(self, blc):
shp = self._shape()
return self._adjust(blc, [0 for x in shp])
# Append trc with shape-1 if shorter than shape.
def _adjustTrc(self, trc):
shp = self._shape()
return self._adjust(trc, [x - 1 for x in shp])
# Append inc with 1 if shorter than shape.
def _adjustInc(self, inc):
shp = self._shape()
return self._adjust(inc, [1 for x in shp])
python-casacore-2.2.1/casacore/measures/ 0000775 0000000 0000000 00000000000 13203113151 0020205 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/measures/__init__.py 0000775 0000000 0000000 00000105604 13203113151 0022327 0 ustar 00root root 0000000 0000000 # __init__.py: Top level .py file for python measures interface
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: __init__.py,v 1.2 2006/12/04 04:01:03 mmarquar Exp $
__all__ = ['is_measure', 'measures']
from ._measures import measures as _measures
import casacore.quanta as dq
import os
if 'MEASURESDATA' in os.environ.keys():
if 'AIPSPATH' not in os.environ.keys():
os.environ['AIPSPATH'] = '%s dummy dummy' % os.environ['MEASURESDATA']
def is_measure(v):
"""
Return if this is a true measures dictionary
:param v: The object to check
"""
if isinstance(v, dict) and v.has_key("type") and v.has_key("m0"):
return True
return False
def _check_valid_offset(self, mtype):
if not off['type'] == mtype:
raise TypeError('Illegal offset type specified.')
class measures(_measures):
"""The measures server object. This should be used to set frame
information and create the various measures and do conversion on them.
The measures types are:
* :meth:`direction`
* :meth:`position`
* :meth:`epoch`
* :meth:`frequency`
* :meth:`doppler`
* :meth:`baseline`
* :meth:`radialvelocity`
* :meth:`uvw`
* :meth:`earthmagnetic`
Typical usage::
from casacore.measures import measures
dm = measures() # create measures server instance
dirmeas = dm.direction()
"""
def __init__(self):
_measures.__init__(self)
self._framestack = {}
def set_data_path(self, pth):
"""Set the location of the measures data directory.
:param pth: The absolute path to the measures data directory.
"""
if os.path.exists(pth):
if not os.path.exists(os.path.join(pth, 'data', 'geodetic')):
raise IOError("The given path doesn't contain a 'data' "
"subdirectory")
os.environ["AIPSPATH"] = "%s dummy dummy" % pth
def measure(self, v, rf, off=None):
"""Create/convert a measure using the frame state set on the measures
server instance (via :meth:`do_frame`)
:param v: The measure to convert
:param rf: The frame reference to convert to
:param off: The optional offset for the measure
"""
if off is None: off = {}
keys = ["m0", "m1", "m2"]
for key in keys:
if key in v:
if dq.is_quantity(v[key]):
v[key] = v[key].to_dict()
return _measures.measure(self, v, rf, off)
def direction(self, rf='', v0='0..', v1='90..', off=None):
"""Defines a direction measure. It has to specify a reference code,
direction quantity values (see introduction for the action on a
scalar quantity with either a vector or scalar value, and when a
vector of quantities is given), and optionally it can specify an
offset, which in itself has to be a direction.
:param rf: reference code string; allowable reference codes are:
J2000 JMEAN JTRUE APP B1950 BMEAN BTRUE GALACTIC HADEC
AZEL SUPERGAL ECLIPTIC MECLIPTIC TECLIPTIC MERCURY VENUS
MARS JUPITER SATURN URANUS NEPTUNE PLUTO MOON SUN COMET.
Note that additional ones may become available. Check with::
dm.list_codes(dm.direction())
:param v0, v1: Direction quantity values should be
longitude (angle) and latitude (angle) or strings
parsable by :func:`~casacore.quanta.quantity`.
None are needed for planets: the frame epoch defines
coordinates. See :func:`~casacore.quanta.quantity` for
possible angle formats.
:param off: an optional offset measure of same type
Example::
>>> dm.direction('j2000','30deg','40deg')
>>> dm.direction('mars')
"""
loc = {'type': 'direction', 'refer': rf}
loc['m0'] = dq.quantity(v0)
loc['m1'] = dq.quantity(v1)
if is_measure(off):
if not off['type'] == "direction":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def position(self, rf='', v0='0..', v1='90..', v2='0m', off=None):
"""Defines a position measure. It has to specify a reference code,
position quantity values (see introduction for the action on a
scalar quantity with either a vector or scalar value, and when a
vector of quantities is given), and optionally it can specify an
offset, which in itself has to be a position.
Note that additional ones may become available. Check with::
dm.list_codes(dm.position())
The position quantity values should be either longitude (angle),
latitude(angle) and height(length); or x,y,z (length). See
:func:`~casacore.quanta.quantity` for possible angle formats.
:param rf: reference code string; Allowable reference
codes are: *WGS84* *ITRF* (World Geodetic System and
International Terrestrial Reference Frame)
:param v0: longitude or x as quantity or string
:param v1: latitude or y as quantity or string
:param v2: height or z as quantity or string
:param off: an optional offset measure of same type
Example::
dm.position('wgs84','30deg','40deg','10m')
dm.observatory('ATCA')
"""
loc = {'type': 'position', 'refer': rf}
loc['m0'] = dq.quantity(v0)
loc['m1'] = dq.quantity(v1)
loc['m2'] = dq.quantity(v2)
if is_measure(off):
if not off['type'] == "position":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def epoch(self, rf='', v0='0.0d', off=None):
"""
Defines an epoch measure. It has to specify a reference code, an epoch
quantity value (see introduction for the action on a scalar quantity
with either a vector or scalar value, and when a vector of quantities
is given), and optionally it can specify an offset, which in itself
has to be an epoch.
:param rf: reference code string; Allowable reference
codes are: *UTC TAI LAST LMST GMST1 GAST UT1 UT2 TDT TCG
TDB TCB*
Note that additional ones may become available. Check with::
dm.list_codes(dm.position())
:param v0: time as quantity or string
:param off: an optional offset measure of same type
"""
loc = {'type': 'epoch', 'refer': rf}
loc['m0'] = dq.quantity(v0)
if is_measure(off):
if not off['type'] == "epoch":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def frequency(self, rf='', v0='0Hz', off=None):
"""Defines a frequency measure. It has to specify a reference code,
frequency quantity value (see introduction for the action on a scalar
quantity with either a vector or scalar value, and when a vector of
quantities is given), and optionally it can specify an offset, which
in itself has to be a frequency.
:param rf: reference code string; Allowable reference
codes are: *REST LSRK LSRD BARY GEO TOPO GALACTO*
Note that additional ones may become available. Check with::
dm.list_codes(dm.frequency())
:param v0: frequency value as quantity or string. The frequency
quantity values should be in one of the recognised units
(examples all give same frequency):
* value with time units: a period (0.5s)
* value as frequency: 2Hz
* value in angular frequency: 720deg/s
* value as length: 149896km
* value as wave number: 4.19169e-8m-1
* value as enery (h.nu): 8.27134e-9ueV
* value as momentum: 4.42044e-42kg.m
:param off: an optional offset measure of same type
"""
loc = {'type': "frequency",
'refer': rf,
'm0': dq.quantity(v0)}
if is_measure(off):
if not off['type'] == "frequency":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def doppler(self, rf='', v0=0.0, off=None):
"""Defines a doppler measure. It has to specify a reference code,
doppler quantity value (see introduction for the action on a scalar
quantity with either a vector or scalar value, and when a vector of
quantities is given), and optionally it can specify an offset, which
in itself has to be a doppler.
:param rf: reference code string; Allowable reference
codes are: *RADIO OPTICAL Z RATIO RELATIVISTIC BETA GAMMA*.
Note that additional ones may become available. Check with::
dm.list_codes(dm.doppler())
:param v0: doppler ratio as quantity, string or float value. It
should be either non-dimensioned to specify a ratio of
the light velocity, or in velocity. (examples all give
same doppler):
:param off: an optional offset measure of same type
Example::
>>> from casacore import quanta
>>> dm.doppler('radio', 0.4)
>>> dm.doppler('radio', '0.4')
>>> dm.doppler('RADIO', quanta.constants['c']*0.4))
"""
if isinstance(v0, float):
v0 = str(v0)
loc = {'type': "doppler",
'refer': rf,
'm0': dq.quantity(v0)}
if is_measure(off):
if not off['type'] == "doppler":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def radialvelocity(self, rf='', v0='0m/s', off=None):
"""Defines a radialvelocity measure. It has to specify a reference
code, radialvelocity quantity value (see introduction for the action
on a scalar quantity with either a vector or scalar value, and when
a vector of quantities is given), and optionally it can specify an
offset, which in itself has to be a radialvelocity.
:param rf: reference code string; Allowable reference
codes are: *LSRK LSRD BARY GEO TOPO GALACTO*
Note that additional ones may become available. Check with::
dm.list_codes(dm.radialvelocity())
:param v0: longitude or x as quantity or string
:param off: an optional offset measure of same type
"""
loc = {'type': "radialvelocity",
'refer': rf,
'm0': dq.quantity(v0)}
if is_measure(off):
if not off['type'] == "radialvelocity":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def baseline(self, rf='', v0='0..', v1='', v2='', off=None):
"""Defines a baselin measure. It has to specify a reference code, uvw
quantity values (see introduction for the action on a scalar quantity
with either a vector or scalar value, and when a vector of quantities
is given), and optionally it can specify an offset, which in itself
has to be a baseline.
:param rf: reference code string; Allowable reference
codes are: *ITRF* and :meth:`direction` codes
Note that additional ones may become available. Check with::
dm.list_codes(dm.baseline())
:param v0: longitude or x as quantity or string
:param v1: latitude or y as quantity or string
:param v2: height or z as quantity or string
:param off: an optional offset measure of same type
"""
loc = {'type': "baseline", 'refer': rf}
loc['m0'] = dq.quantity(v0)
loc['m1'] = dq.quantity(v1)
loc['m2'] = dq.quantity(v2)
if is_measure(off):
if not off['type'] == "doppler":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def uvw(self, rf='', v0='0..', v1='', v2='', off=None):
"""Defines a uvw measure. It has to specify a reference code, uvw
quantity values (see introduction for the action on a scalar quantity
with either a vector or scalar value, and when a vector of quantities
is given), and optionally it can specify an offset, which in itself
has to be a uvw.
:param rf: reference code string; Allowable reference
codes are: *ITRF* and :meth:`direction` codes
Note that additional ones may become available. Check with::
dm.list_codes(dm.uvw())
:param v0: longitude or x as quantity or string
:param v1: latitude or y as quantity or string
:param v2: height or z as quantity or string
:param off: an optional offset measure of same type
"""
loc = {'type': "uvw", 'refer': rf}
loc['m0'] = dq.quantity(v0)
loc['m1'] = dq.quantity(v1)
loc['m2'] = dq.quantity(v2)
if is_measure(off):
if not off['type'] == "uvw":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def earthmagnetic(self, rf='', v0='0G', v1='0..', v2='90..', off=None):
"""Defines an earthmagnetic measure. It needs a reference code,
earthmagnetic quantity values (see introduction for the action on a
scalar quantity with either a vector or scalar value, and when a
vector of quantities is given) if the reference code is not for a
model, and optionally it can specify an offset, which in itself has
to be a earthmagnetic. In general you specify a model (*IGRF* is the
default and the only one known) and convert it to an explicit field.
(See http://fdd.gsfc.nasa.gov/IGRF.html for information on the
International Geomagnetic Reference Field). The earthmagnetic quantity
values should be either longitude (angle), latitude(angle) and
length(field strength); or x,y,z (field).
See :func:`~casacore.quanta.quantity` for possible angle formats.
:param rf: reference code string; Allowable reference
codes are: *IGRF*
:param v0: longitude or x as quantity or string
:param v1: latitude or y as quantity or string
:param v2: height or z as quantity or string
:param off: an optional offset measure of same type
"""
loc = {'type': "earthmagnetic", 'refer': rf}
loc['m0'] = dq.quantity(v0)
loc['m1'] = dq.quantity(v1)
loc['m2'] = dq.quantity(v2)
if is_measure(off):
if not off['type'] == "earthmagnetic":
raise TypeError('Illegal offset type specified.')
loc["offset"] = off
return self.measure(loc, rf)
def tofrequency(self, rf, v0, rfq):
"""Convert a Doppler type value (e.g. in radio mode) to a
frequency. The type of frequency (e.g. LSRK) and a rest frequency
(either as a frequency quantity (e.g. ``dm.constants('HI'))`` or
a frequency measure (e.g. ``dm.frequency('rest','5100MHz'))`` should
be specified.
:param rf: frequency reference code (see :meth:`frequency`)
:param v0: a doppler measure
:param rfq: frequency measure or quantity
Example::
dop = dm.doppler('radio',0.4)
freq = dm.tofrequency('lsrk', dop, dm.constants('HI'))
"""
if is_measure(rfq) and rfq['type'] == 'frequency':
rfq = dq.quantity(rfq['m0'])
elif isinstance(rfq, str):
rfq = dq.quantity(rfq)
if is_measure(v0) and v0['type'] == 'doppler' \
and dq.is_quantity(rfq) \
and rfq.conforms(dq.quantity('Hz')):
return self.doptofreq(v0, rf, rfq)
else:
raise TypeError('Illegal Doppler or rest frequency specified')
to_frequency = tofrequency
def torestfrequency(self, f0, d0):
"""Convert a frequency measure and a doppler measure (e.g.
obtained from another spectral line with a known rest frequency) to
a rest frequency.
:param f0: frequency reference code (see :meth:`frequency`)
:param v0: a doppler measure
Example::
dp = dm.doppler('radio', '2196.24984km/s') # a measured doppler speed
f = dm.frequency('lsrk','1410MHz') # a measured frequency
dm.torestfrequency(f, dp) # the corresponding rest frequency
"""
if is_measure(f0) and f0['type'] == 'frequency' \
and is_measure(d0) and d0['type'] == 'doppler':
return self.torest(f0, d0)
else:
raise TypeError('Illegal Doppler or rest frequency specified')
to_restfrequency = torestfrequency
def todoppler(self, rf, v0, rfq):
"""Convert a radialvelocity measure or a frequency measure to a
doppler measure. In the case of a frequency, a rest frequency has
to be specified. The type of doppler wanted (e.g. *RADIO*) has to be
specified.
:param rf: doppler reference code (see :meth:`doppler`)
:param v0: a radialvelocity or frequency measure
:param rfq: frequency measure or quantity
Example::
f = dm.frequency('lsrk','1410MHz') # specify a frequency
dm.todoppler('radio', f, dm.constants('HI')) # give doppler, using HI rest
"""
if is_measure(rfq) and rfq['type'] == 'frequency':
rfq = dq.quantity(rfq['m0'])
elif isinstance(rfq, str):
rfq = dq.quantity(rfq)
if is_measure(v0):
if v0['type'] == 'radialvelocity':
return self.todop(v0, dq.quantity(1., 'Hz'))
elif v0['type'] == 'frequency' and dq.is_quantity(rfq) \
and rfq.conforms(dq.quantity('Hz')):
return self.todop(v0, rfq)
else:
raise TypeError('Illegal Doppler or rest frequency specified')
else:
raise TypeError('Illegal Frequency specified')
to_doppler = todoppler
def toradialvelocity(self, rf, v0):
"""Convert a Doppler type value (e.g. in radio mode) to a real
radialvelocity. The type of velocity (e.g. *LSRK*) should be specified
:param rf: radialvelocity reference code (see :meth:`radialvelocity`)
:param v0: a doppler measure
Example::
a = dm.doppler('radio',0.4)
dm.toradialvelocity('topo',a)
"""
if is_measure(v0) and v0['type'] == 'doppler':
return self.doptorv(rf, v0)
else:
raise TypeError('Illegal Doppler specified')
to_radialvelocity = toradialvelocity
def touvw(self, v):
"""Calculates a uvw measure from a baseline. The baseline can consist
of a vector of actual baseline positions. Note that the baseline does
not have to be a proper baseline, but can be a series of positions
(to call positions baselines see :meth:`asbaseline` ) for speed
reasons: operations are linear and can be done on positions, which
are converted to baseline values at the end (with :meth:`expand` ).
Whatever the reference code of the baseline, the returned uvw will be
given in J2000. If the dot argument is given, that variable will be
filled with a quantity array consisting of the time derivative of the
uvw (note that only the sidereal rate is taken into account; not
precession, earth tides and similar variations, which are much
smaller). If the xyz variable is given, it will be filled with the
quantity values of the uvw measure.
The values of the input baselines can be given as a quantity
vector per x, y or z value.
uvw coordinates are calculated for a certain direction in the sky
hence the frame has to contain the direction for the calculation to
work. Since the baseline and the sky rotate with respect of each
other, the time should be specified as well.
Example::
>>> dm.do_frame(dm.observatory('atca'))
>>> dm.do_frame(dm.source('1934-638'))
>>> dm.do_frame(dm.epoch('utc', 'today'))
>>> b = dm.baseline('itrf', '10m', '20m', '30m')
"""
if is_measure(v) and v['type'] == 'baseline':
m = _measures.uvw(self, v)
m['xyz'] = dq.quantity(m['xyz'])
m['dot'] = dq.quantity(m['dot'])
return m
else:
raise TypeError('Illegal Baseline specified')
to_uvw = touvw
def expand(self, v):
"""Calculates the differences between a series of given measure values:
it calculates baseline values from position values.
:params v: a measure (of type 'baseline', 'position' or 'uvw')
:returns: a `dict` with the value for key `measures` being a measure
and the value for key `xyz` a quantity containing the
differences.
Example::
>>> from casacore.quanta import quantity
>>> x = quantity([10,50],'m')
>>> y = quantity([20,100],'m')
>>> z = quantity([30,150],'m')
>>> sb = dm.baseline('itrf', x, y, z)
>>> out = dm.expand(sb)
>>> print out['xyz']
[40.000000000000014, 80.0, 120.0] m
"""
if not is_measure(v) or v['type'] not in ['baseline',
'position', 'uvw']:
raise TypeError("Can only expand baselines, positions, or uvw")
vw = v.copy()
vw['type'] = "uvw"
vw['refer'] = "J2000"
outm = _measures.expand(self, vw)
outm['xyz'] = dq.quantity(outm['xyz'])
outm['measure']['type'] = v['type']
outm['measure']['refer'] = v['refer']
return outm
def asbaseline(self, pos):
"""Convert a position measure into a baseline measure. No actual
baseline is calculated, since operations can be done on positions,
with subtractions to obtain baselines at a later stage.
:param pos: a position measure
:returns: a baseline measure
"""
if not is_measure(pos) or pos['type'] not in ['position', 'baseline']:
raise TypeError('Argument is not a position/baseline measure')
if pos['type'] == 'position':
loc = self.measure(pos, 'itrf')
loc['type'] = 'baseline'
return self.measure(loc, 'j2000')
return pos
as_baseline = asbaseline
def getvalue(self, v):
"""
Return a list of quantities making up the measures' value.
:param v: a measure
"""
if not is_measure(v):
raise TypeError('Incorrect input type for getvalue()')
import re
rx = re.compile("m\d+")
out = []
keys = v.keys()[:]
keys.sort()
for key in keys:
if re.match(rx, key):
out.append(dq.quantity(v.get(key)))
return out
get_value = getvalue
def get_type(self, m):
"""Get the type of the measure.
:param m: a measure (dictionary)
:rtype: string
"""
if is_measure(m):
return m["type"]
else:
raise TypeError("Argument is not a measure")
def get_ref(self, m):
"""Get the reference frame of the measure.
:param m: a measure (dictionary)
:rtype: string
"""
if is_measure(m):
return m["refer"]
else:
raise TypeError("Argument is not a measure")
def get_offset(self, m):
"""Get the offset measure.
:param m: a measure (dictionary)
:rtype: a measure
"""
if is_measure(m):
return m.get("offset", None)
else:
raise TypeError("Argument is not a measure")
def doframe(self, v):
"""This method will set the measure specified as part of a frame.
If conversion from one type to another is necessary (with the measure
function), the following frames should be set if one of the reference
types involved in the conversion is as in the following lists:
**Epoch**
* UTC
* TAI
* LAST - position
* LMST - position
* GMST1
* GAST
* UT1
* UT2
* TDT
* TCG
* TDB
* TCD
**Direction**
* J2000
* JMEAN - epoch
* JTRUE - epoch
* APP - epoch
* B1950
* BMEAN - epoch
* BTRUE - epoch
* GALACTIC
* HADEC - epoch, position
* AZEL - epoch, position
* SUPERGALACTIC
* ECLIPTIC
* MECLIPTIC - epoch
* TECLIPTIC - epoch
* PLANET - epoch, [position]
**Position**
* WGS84
* ITRF
**Radial Velocity**
* LSRK - direction
* LSRD - direction
* BARY - direction
* GEO - direction, epoch
* TOPO - direction, epoch, position
* GALACTO - direction
*
**Doppler**
* RADIO
* OPTICAL
* Z
* RATIO
* RELATIVISTIC
* BETA
* GAMMA
*
**Frequency**
* REST - direction, radialvelocity
* LSRK - direction
* LSRD - direction
* BARY - direction
* GEO - direction, epoch
* TOPO - direction, epoch, position
* GALACTO
"""
if not is_measure(v):
raise TypeError('Argument is not a measure')
if (v["type"] == "frequency" and v["refer"].lower() == "rest") \
or _measures.doframe(self, v):
self._framestack[v["type"]] = v
return True
return False
do_frame = doframe
def _fillnow(self):
if not "epoch" in self._framestack \
or not is_measure(self._framestack["epoch"]):
self.frame_now()
def _getwhere(self):
if not self._framestack.has_key("position") \
or not is_measure(self._framestack["position"]):
raise RuntimeError("Can't find position frame")
return self._framestack["position"]
def framenow(self):
"""Set the time (epoch) frame to the current time and day."""
self.do_frame(self.epoch("UTC", "today"))
frame_now = framenow
def rise(self, crd, ev='5deg'):
"""This method will give the rise/set hour-angles of a source. It
needs the position in the frame, and a time. If the latter is not
set, the current time will be used.
:param crd: a direction measure
:param ev: the elevation limit as a quantity or string
:returns: `dict` with rise and set sidereal time quantities or a 2
strings "below" or "above"
"""
if not is_measure(crd):
raise TypeError('No rise/set coordinates specified')
ps = self._getwhere()
self._fillnow()
hd = self.measure(crd, "hadec")
c = self.measure(crd, "app")
evq = dq.quantity(ev)
hdm1 = dq.quantity(hd["m1"])
psm1 = dq.quantity(ps["m1"])
ct = (dq.sin(dq.quantity(ev)) - (dq.sin(hdm1) * dq.sin(psm1))) \
/ (dq.cos(hdm1) * dq.cos(psm1))
if ct.get_value() >= 1:
return {'rise': 'below', 'set': 'below'}
if ct.get_value() <= -1:
return {'rise': 'above', 'set': 'above'}
a = dq.acos(ct)
return dict(rise=dq.quantity(c["m0"]).norm(0) - a,
set=dq.quantity(c["m0"]).norm(0) + a)
def riseset(self, crd, ev="5deg"):
"""This will give the rise/set times of a source. It needs the
position in the frame, and a time. If the latter is not set, the
current time will be used.
:param crd: a direction measure
:param ev: the elevation limit as a quantity or string
:returns: The returned value is a `dict` with a
'solved' key, which is `False` if the source is always
below or above the horizon. In that case the rise and set
fields will all have a string value. The `dict` also returns
a rise and set `dict`, with 'last' and 'utc' keys showing
the rise and set times as epochs.
"""
a = self.rise(crd, ev)
if isinstance(a['rise'], str):
return {"rise": {"last": a[0], "utc": a[0]},
"set": {"last": a[1], "utc": a[1]},
"solved": False}
ofe = self.measure(self._framestack["epoch"], "utc")
if not is_measure(ofe):
ofe = self.epoch('utc', 'today')
x = a.copy()
for k in x:
x[k] = self.measure(
self.epoch("last",
a[k].totime(),
off=self.epoch("r_utc",
(dq.quantity(ofe["m0"])
+ dq.quantity("0.5d")
))
),
"utc")
return {"rise": {"last": self.epoch("last",
a["rise"].totime()),
"utc": x["rise"]},
"set": {"last": self.epoch("last",
a["set"].totime()),
"utc": x["set"]},
"solved": True
}
def observatory(self, name):
"""Get a (position) measure for the given obervatory.
:param name: the name of the observatory. At the time of
writing the following observatories are recognised (but
check :meth:`get_observatories`): *ALMA ATCA BIMA CLRO
DRAO DWL GB JCMT MOPRA NRAO12M PKS VLA WSRT*
:returns: a position measure
"""
return _measures.observatory(self, name.upper())
def get_observatories(self):
"""Return a list of known observatory names, which can be used as input
to :meth:`observatory`.
:rtype: list of strings
"""
return self.obslist()
def line(self, name):
"""Get a (frequency) measure for the given spectral line name.
:param name: the name of the spectral line. Minimum match applies.
At the time of writing the following are recognised (but
check :meth:`get_lines`): *C109A CI CII166A DI H107A
H110A H138B H166A H240A H272A H2CO HE110A HE138B HI
OH1612 OH1665 OH1667 OH1720 CO115271 H2O22235 SiO86847
CO230538*
:returns: a frequency measure
"""
return _measures.line(self, name)
def get_lines(self):
"""Return a list of known spectral line names, which can be used as
input to :meth:`line`.
:rtype: list of strings
"""
return self.linelist()
def source(self, name):
"""Get a (direction) measure for the given atsronomical source.
:param name: the name of the source. Minimum match applies.
Check :meth:`get_sources` for the list of known sources
:returns: a frequency measure
Example::
>>> print dm.source('1936-6')
{'m0': {'unit': 'rad', 'value': -1.1285176426372401},
'm1': {'unit': 'rad', 'value': -1.0854059868642842},
'refer': 'ICRS',
'type': 'direction'}
"""
return _measures.source(self, name)
def get_sources(self):
"""Return a list of known sources names, which can be used as
input to :meth:`source`.
:rtype: list of strings
"""
return self.srclist()
# alias
def list_codes(self, m):
"""Get the known reference codes for a specified measure type. It
will return a `dict` with two keys. The first is a string list
of all normal codes; the second a string list (maybe empty) with all
extra codes (like planets).
:param m: the measures with the type to get codes for
"""
return _measures.alltyp(self, m)
def posangle(self, m0, m1):
"""
This method will give the position angle from a direction to another
i.e. the angle in a direction between the direction to the North
pole and the other direction.
:param m0: a measure
:param m1: another measure
Example::
>>> a = dm.direction('j2000','0deg','70deg')
>>> b = dm.direction('j2000','0deg','80deg')
>>> print dm.posangle(a,b)
-0.0 deg
"""
return _measures.posangle(self, m0, m1)
def separation(self, m0, m1):
"""
This method will give the separation of a direction from another as
an angle.
:param m0: a measure
:param m1: another measure
Example::
>>> a = dm.direction('j2000','0deg','70deg')
>>> b = dm.direction('j2000','0deg','80deg')
>>> print dm.separation(a,b)
10.0 deg
"""
return _measures.separation(self, m0, m1)
## def show(v, refcode=True):
## z = ""
## if is_measure(v):
## x = dm.gettype(v)
## y = dm.getvalue(v)
## if x.startswith("epo"):
## z = dq.form.dtime(y[0])
## else:
## return ""
## if refcode:
## return [z, dm.getref(v)]
## return z
python-casacore-2.2.1/casacore/quanta/ 0000775 0000000 0000000 00000000000 13203113151 0017652 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/quanta/__init__.py 0000664 0000000 0000000 00000000264 13203113151 0021765 0 ustar 00root root 0000000 0000000 from ._quanta import *
from .quantity import quantity, is_quantity
constants = constants()
units = units()
prefixes = prefixes()
del Quantity, QuantVec, from_string, from_dict_v
python-casacore-2.2.1/casacore/quanta/quantity.py 0000664 0000000 0000000 00000003546 13203113151 0022112 0 ustar 00root root 0000000 0000000 from ._quanta import QuantVec
from ._quanta import Quantity
from ._quanta import from_string, from_dict, from_dict_v
def is_quantity(q):
"""Indicate whether the object is a valid quantity"""
return isinstance(q, QuantVec) or isinstance(q, Quantity)
# Quantity returns new Quantities, so we need to insert these
# functions into Quantity
def new_get_value(quant, *args):
val = QuantVec._get_value(quant, *args)
if len(val) == 1:
return val[0]
else:
return val
QuantVec.get_value = new_get_value
def to_string(quant, fmt="%0.5f"):
val = quant.get_value()
if hasattr(val, "__len__"):
fmt = "[" + ", ".join([fmt % i for i in val]) + "] %s"
return fmt % quant.get_unit()
fmt += " %s"
return fmt % (val, quant.get_unit())
QuantVec.to_string = to_string
Quantity.to_string = to_string
QuantVec.__str__ = to_string
Quantity.__str__ = to_string
# QuantVec.__repr__ = to_string
# Quantity.__repr__ = to_string
def quantity(*args):
"""Create a quantity. This can be from a scalar or vector.
Example::
q1 = quantity(1.0, "km/s")
q2 = quantity("1km/s")
q1 = quantity([1.0,2.0], "km/s")
"""
if len(args) == 1:
if isinstance(args[0], str):
# use copy constructor to create quantity from string
return Quantity(from_string(args[0]))
elif isinstance(args[0], dict):
if hasattr(args[0]["value"], "__len__"):
return QuantVec(from_dict_v(args[0]))
else:
return Quantity(from_dict(args[0]))
elif isinstance(args[0], Quantity) or isinstance(args[0], QuantVec):
return args[0]
else:
raise TypeError("Invalid argument type for")
else:
if hasattr(args[0], "__len__"):
return QuantVec(*args)
else:
return Quantity(*args)
python-casacore-2.2.1/casacore/six.py 0000664 0000000 0000000 00000066532 13203113151 0017552 0 ustar 00root root 0000000 0000000 """Utilities for writing code that runs on Python 2 and 3"""
# Copyright (c) 2010-2014 Benjamin Peterson
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
from __future__ import absolute_import
import functools
import itertools
import operator
import sys
import types
__author__ = "Benjamin Peterson "
__version__ = "1.8.0"
# Useful for very coarse version differentiation.
PY2 = sys.version_info[0] == 2
PY3 = sys.version_info[0] == 3
if PY3:
string_types = str,
integer_types = int,
class_types = type,
text_type = str
binary_type = bytes
MAXSIZE = sys.maxsize
else:
string_types = basestring,
integer_types = (int, long)
class_types = (type, types.ClassType)
text_type = unicode
binary_type = str
if sys.platform.startswith("java"):
# Jython always uses 32 bits.
MAXSIZE = int((1 << 31) - 1)
else:
# It's possible to have sizeof(long) != sizeof(Py_ssize_t).
class X(object):
def __len__(self):
return 1 << 31
try:
len(X())
except OverflowError:
# 32-bit
MAXSIZE = int((1 << 31) - 1)
else:
# 64-bit
MAXSIZE = int((1 << 63) - 1)
del X
def _add_doc(func, doc):
"""Add documentation to a function."""
func.__doc__ = doc
def _import_module(name):
"""Import module, returning the module after the last dot."""
__import__(name)
return sys.modules[name]
class _LazyDescr(object):
def __init__(self, name):
self.name = name
def __get__(self, obj, tp):
result = self._resolve()
setattr(obj, self.name, result) # Invokes __set__.
try:
# This is a bit ugly, but it avoids running this again by
# removing this descriptor.
delattr(obj.__class__, self.name)
except AttributeError:
pass
return result
class MovedModule(_LazyDescr):
def __init__(self, name, old, new=None):
super(MovedModule, self).__init__(name)
if PY3:
if new is None:
new = name
self.mod = new
else:
self.mod = old
def _resolve(self):
return _import_module(self.mod)
def __getattr__(self, attr):
_module = self._resolve()
value = getattr(_module, attr)
setattr(self, attr, value)
return value
class _LazyModule(types.ModuleType):
def __init__(self, name):
super(_LazyModule, self).__init__(name)
self.__doc__ = self.__class__.__doc__
def __dir__(self):
attrs = ["__doc__", "__name__"]
attrs += [attr.name for attr in self._moved_attributes]
return attrs
# Subclasses should override this
_moved_attributes = []
class MovedAttribute(_LazyDescr):
def __init__(self, name, old_mod, new_mod, old_attr=None, new_attr=None):
super(MovedAttribute, self).__init__(name)
if PY3:
if new_mod is None:
new_mod = name
self.mod = new_mod
if new_attr is None:
if old_attr is None:
new_attr = name
else:
new_attr = old_attr
self.attr = new_attr
else:
self.mod = old_mod
if old_attr is None:
old_attr = name
self.attr = old_attr
def _resolve(self):
module = _import_module(self.mod)
return getattr(module, self.attr)
class _SixMetaPathImporter(object):
"""
A meta path importer to import six.moves and its submodules.
This class implements a PEP302 finder and loader. It should be compatible
with Python 2.5 and all existing versions of Python3
"""
def __init__(self, six_module_name):
self.name = six_module_name
self.known_modules = {}
def _add_module(self, mod, *fullnames):
for fullname in fullnames:
self.known_modules[self.name + "." + fullname] = mod
def _get_module(self, fullname):
return self.known_modules[self.name + "." + fullname]
def find_module(self, fullname, path=None):
if fullname in self.known_modules:
return self
return None
def __get_module(self, fullname):
try:
return self.known_modules[fullname]
except KeyError:
raise ImportError("This loader does not know module " + fullname)
def load_module(self, fullname):
try:
# in case of a reload
return sys.modules[fullname]
except KeyError:
pass
mod = self.__get_module(fullname)
if isinstance(mod, MovedModule):
mod = mod._resolve()
else:
mod.__loader__ = self
sys.modules[fullname] = mod
return mod
def is_package(self, fullname):
"""
Return true, if the named module is a package.
We need this method to get correct spec objects with
Python 3.4 (see PEP451)
"""
return hasattr(self.__get_module(fullname), "__path__")
def get_code(self, fullname):
"""Return None
Required, if is_package is implemented"""
self.__get_module(fullname) # eventually raises ImportError
return None
get_source = get_code # same as get_code
_importer = _SixMetaPathImporter(__name__)
class _MovedItems(_LazyModule):
"""Lazy loading of moved objects"""
__path__ = [] # mark as package
_moved_attributes = [
MovedAttribute("cStringIO", "cStringIO", "io", "StringIO"),
MovedAttribute("filter", "itertools", "builtins", "ifilter", "filter"),
MovedAttribute("filterfalse", "itertools", "itertools", "ifilterfalse", "filterfalse"),
MovedAttribute("input", "__builtin__", "builtins", "raw_input", "input"),
MovedAttribute("intern", "__builtin__", "sys"),
MovedAttribute("map", "itertools", "builtins", "imap", "map"),
MovedAttribute("range", "__builtin__", "builtins", "xrange", "range"),
MovedAttribute("reload_module", "__builtin__", "imp", "reload"),
MovedAttribute("reduce", "__builtin__", "functools"),
MovedAttribute("shlex_quote", "pipes", "shlex", "quote"),
MovedAttribute("StringIO", "StringIO", "io"),
MovedAttribute("UserDict", "UserDict", "collections"),
MovedAttribute("UserList", "UserList", "collections"),
MovedAttribute("UserString", "UserString", "collections"),
MovedAttribute("xrange", "__builtin__", "builtins", "xrange", "range"),
MovedAttribute("zip", "itertools", "builtins", "izip", "zip"),
MovedAttribute("zip_longest", "itertools", "itertools", "izip_longest", "zip_longest"),
MovedModule("builtins", "__builtin__"),
MovedModule("configparser", "ConfigParser"),
MovedModule("copyreg", "copy_reg"),
MovedModule("dbm_gnu", "gdbm", "dbm.gnu"),
MovedModule("_dummy_thread", "dummy_thread", "_dummy_thread"),
MovedModule("http_cookiejar", "cookielib", "http.cookiejar"),
MovedModule("http_cookies", "Cookie", "http.cookies"),
MovedModule("html_entities", "htmlentitydefs", "html.entities"),
MovedModule("html_parser", "HTMLParser", "html.parser"),
MovedModule("http_client", "httplib", "http.client"),
MovedModule("email_mime_multipart", "email.MIMEMultipart", "email.mime.multipart"),
MovedModule("email_mime_nonmultipart", "email.MIMENonMultipart", "email.mime.nonmultipart"),
MovedModule("email_mime_text", "email.MIMEText", "email.mime.text"),
MovedModule("email_mime_base", "email.MIMEBase", "email.mime.base"),
MovedModule("BaseHTTPServer", "BaseHTTPServer", "http.server"),
MovedModule("CGIHTTPServer", "CGIHTTPServer", "http.server"),
MovedModule("SimpleHTTPServer", "SimpleHTTPServer", "http.server"),
MovedModule("cPickle", "cPickle", "pickle"),
MovedModule("queue", "Queue"),
MovedModule("reprlib", "repr"),
MovedModule("socketserver", "SocketServer"),
MovedModule("_thread", "thread", "_thread"),
MovedModule("tkinter", "Tkinter"),
MovedModule("tkinter_dialog", "Dialog", "tkinter.dialog"),
MovedModule("tkinter_filedialog", "FileDialog", "tkinter.filedialog"),
MovedModule("tkinter_scrolledtext", "ScrolledText", "tkinter.scrolledtext"),
MovedModule("tkinter_simpledialog", "SimpleDialog", "tkinter.simpledialog"),
MovedModule("tkinter_tix", "Tix", "tkinter.tix"),
MovedModule("tkinter_ttk", "ttk", "tkinter.ttk"),
MovedModule("tkinter_constants", "Tkconstants", "tkinter.constants"),
MovedModule("tkinter_dnd", "Tkdnd", "tkinter.dnd"),
MovedModule("tkinter_colorchooser", "tkColorChooser",
"tkinter.colorchooser"),
MovedModule("tkinter_commondialog", "tkCommonDialog",
"tkinter.commondialog"),
MovedModule("tkinter_tkfiledialog", "tkFileDialog", "tkinter.filedialog"),
MovedModule("tkinter_font", "tkFont", "tkinter.font"),
MovedModule("tkinter_messagebox", "tkMessageBox", "tkinter.messagebox"),
MovedModule("tkinter_tksimpledialog", "tkSimpleDialog",
"tkinter.simpledialog"),
MovedModule("urllib_parse", __name__ + ".moves.urllib_parse", "urllib.parse"),
MovedModule("urllib_error", __name__ + ".moves.urllib_error", "urllib.error"),
MovedModule("urllib", __name__ + ".moves.urllib", __name__ + ".moves.urllib"),
MovedModule("urllib_robotparser", "robotparser", "urllib.robotparser"),
MovedModule("xmlrpc_client", "xmlrpclib", "xmlrpc.client"),
MovedModule("xmlrpc_server", "SimpleXMLRPCServer", "xmlrpc.server"),
MovedModule("winreg", "_winreg"),
]
for attr in _moved_attributes:
setattr(_MovedItems, attr.name, attr)
if isinstance(attr, MovedModule):
_importer._add_module(attr, "moves." + attr.name)
del attr
_MovedItems._moved_attributes = _moved_attributes
moves = _MovedItems(__name__ + ".moves")
_importer._add_module(moves, "moves")
class Module_six_moves_urllib_parse(_LazyModule):
"""Lazy loading of moved objects in six.moves.urllib_parse"""
_urllib_parse_moved_attributes = [
MovedAttribute("ParseResult", "urlparse", "urllib.parse"),
MovedAttribute("SplitResult", "urlparse", "urllib.parse"),
MovedAttribute("parse_qs", "urlparse", "urllib.parse"),
MovedAttribute("parse_qsl", "urlparse", "urllib.parse"),
MovedAttribute("urldefrag", "urlparse", "urllib.parse"),
MovedAttribute("urljoin", "urlparse", "urllib.parse"),
MovedAttribute("urlparse", "urlparse", "urllib.parse"),
MovedAttribute("urlsplit", "urlparse", "urllib.parse"),
MovedAttribute("urlunparse", "urlparse", "urllib.parse"),
MovedAttribute("urlunsplit", "urlparse", "urllib.parse"),
MovedAttribute("quote", "urllib", "urllib.parse"),
MovedAttribute("quote_plus", "urllib", "urllib.parse"),
MovedAttribute("unquote", "urllib", "urllib.parse"),
MovedAttribute("unquote_plus", "urllib", "urllib.parse"),
MovedAttribute("urlencode", "urllib", "urllib.parse"),
MovedAttribute("splitquery", "urllib", "urllib.parse"),
MovedAttribute("splittag", "urllib", "urllib.parse"),
MovedAttribute("splituser", "urllib", "urllib.parse"),
MovedAttribute("uses_fragment", "urlparse", "urllib.parse"),
MovedAttribute("uses_netloc", "urlparse", "urllib.parse"),
MovedAttribute("uses_params", "urlparse", "urllib.parse"),
MovedAttribute("uses_query", "urlparse", "urllib.parse"),
MovedAttribute("uses_relative", "urlparse", "urllib.parse"),
]
for attr in _urllib_parse_moved_attributes:
setattr(Module_six_moves_urllib_parse, attr.name, attr)
del attr
Module_six_moves_urllib_parse._moved_attributes = _urllib_parse_moved_attributes
_importer._add_module(Module_six_moves_urllib_parse(__name__ + ".moves.urllib_parse"),
"moves.urllib_parse", "moves.urllib.parse")
class Module_six_moves_urllib_error(_LazyModule):
"""Lazy loading of moved objects in six.moves.urllib_error"""
_urllib_error_moved_attributes = [
MovedAttribute("URLError", "urllib2", "urllib.error"),
MovedAttribute("HTTPError", "urllib2", "urllib.error"),
MovedAttribute("ContentTooShortError", "urllib", "urllib.error"),
]
for attr in _urllib_error_moved_attributes:
setattr(Module_six_moves_urllib_error, attr.name, attr)
del attr
Module_six_moves_urllib_error._moved_attributes = _urllib_error_moved_attributes
_importer._add_module(Module_six_moves_urllib_error(__name__ + ".moves.urllib.error"),
"moves.urllib_error", "moves.urllib.error")
class Module_six_moves_urllib_request(_LazyModule):
"""Lazy loading of moved objects in six.moves.urllib_request"""
_urllib_request_moved_attributes = [
MovedAttribute("urlopen", "urllib2", "urllib.request"),
MovedAttribute("install_opener", "urllib2", "urllib.request"),
MovedAttribute("build_opener", "urllib2", "urllib.request"),
MovedAttribute("pathname2url", "urllib", "urllib.request"),
MovedAttribute("url2pathname", "urllib", "urllib.request"),
MovedAttribute("getproxies", "urllib", "urllib.request"),
MovedAttribute("Request", "urllib2", "urllib.request"),
MovedAttribute("OpenerDirector", "urllib2", "urllib.request"),
MovedAttribute("HTTPDefaultErrorHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPRedirectHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPCookieProcessor", "urllib2", "urllib.request"),
MovedAttribute("ProxyHandler", "urllib2", "urllib.request"),
MovedAttribute("BaseHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPPasswordMgr", "urllib2", "urllib.request"),
MovedAttribute("HTTPPasswordMgrWithDefaultRealm", "urllib2", "urllib.request"),
MovedAttribute("AbstractBasicAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPBasicAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("ProxyBasicAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("AbstractDigestAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPDigestAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("ProxyDigestAuthHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPSHandler", "urllib2", "urllib.request"),
MovedAttribute("FileHandler", "urllib2", "urllib.request"),
MovedAttribute("FTPHandler", "urllib2", "urllib.request"),
MovedAttribute("CacheFTPHandler", "urllib2", "urllib.request"),
MovedAttribute("UnknownHandler", "urllib2", "urllib.request"),
MovedAttribute("HTTPErrorProcessor", "urllib2", "urllib.request"),
MovedAttribute("urlretrieve", "urllib", "urllib.request"),
MovedAttribute("urlcleanup", "urllib", "urllib.request"),
MovedAttribute("URLopener", "urllib", "urllib.request"),
MovedAttribute("FancyURLopener", "urllib", "urllib.request"),
MovedAttribute("proxy_bypass", "urllib", "urllib.request"),
]
for attr in _urllib_request_moved_attributes:
setattr(Module_six_moves_urllib_request, attr.name, attr)
del attr
Module_six_moves_urllib_request._moved_attributes = _urllib_request_moved_attributes
_importer._add_module(Module_six_moves_urllib_request(__name__ + ".moves.urllib.request"),
"moves.urllib_request", "moves.urllib.request")
class Module_six_moves_urllib_response(_LazyModule):
"""Lazy loading of moved objects in six.moves.urllib_response"""
_urllib_response_moved_attributes = [
MovedAttribute("addbase", "urllib", "urllib.response"),
MovedAttribute("addclosehook", "urllib", "urllib.response"),
MovedAttribute("addinfo", "urllib", "urllib.response"),
MovedAttribute("addinfourl", "urllib", "urllib.response"),
]
for attr in _urllib_response_moved_attributes:
setattr(Module_six_moves_urllib_response, attr.name, attr)
del attr
Module_six_moves_urllib_response._moved_attributes = _urllib_response_moved_attributes
_importer._add_module(Module_six_moves_urllib_response(__name__ + ".moves.urllib.response"),
"moves.urllib_response", "moves.urllib.response")
class Module_six_moves_urllib_robotparser(_LazyModule):
"""Lazy loading of moved objects in six.moves.urllib_robotparser"""
_urllib_robotparser_moved_attributes = [
MovedAttribute("RobotFileParser", "robotparser", "urllib.robotparser"),
]
for attr in _urllib_robotparser_moved_attributes:
setattr(Module_six_moves_urllib_robotparser, attr.name, attr)
del attr
Module_six_moves_urllib_robotparser._moved_attributes = _urllib_robotparser_moved_attributes
_importer._add_module(Module_six_moves_urllib_robotparser(__name__ + ".moves.urllib.robotparser"),
"moves.urllib_robotparser", "moves.urllib.robotparser")
class Module_six_moves_urllib(types.ModuleType):
"""Create a six.moves.urllib namespace that resembles the Python 3 namespace"""
__path__ = [] # mark as package
parse = _importer._get_module("moves.urllib_parse")
error = _importer._get_module("moves.urllib_error")
request = _importer._get_module("moves.urllib_request")
response = _importer._get_module("moves.urllib_response")
robotparser = _importer._get_module("moves.urllib_robotparser")
def __dir__(self):
return ['parse', 'error', 'request', 'response', 'robotparser']
_importer._add_module(Module_six_moves_urllib(__name__ + ".moves.urllib"),
"moves.urllib")
def add_move(move):
"""Add an item to six.moves."""
setattr(_MovedItems, move.name, move)
def remove_move(name):
"""Remove item from six.moves."""
try:
delattr(_MovedItems, name)
except AttributeError:
try:
del moves.__dict__[name]
except KeyError:
raise AttributeError("no such move, %r" % (name,))
if PY3:
_meth_func = "__func__"
_meth_self = "__self__"
_func_closure = "__closure__"
_func_code = "__code__"
_func_defaults = "__defaults__"
_func_globals = "__globals__"
else:
_meth_func = "im_func"
_meth_self = "im_self"
_func_closure = "func_closure"
_func_code = "func_code"
_func_defaults = "func_defaults"
_func_globals = "func_globals"
try:
advance_iterator = next
except NameError:
def advance_iterator(it):
return it.next()
next = advance_iterator
try:
callable = callable
except NameError:
def callable(obj):
return any("__call__" in klass.__dict__ for klass in type(obj).__mro__)
if PY3:
def get_unbound_function(unbound):
return unbound
create_bound_method = types.MethodType
Iterator = object
else:
def get_unbound_function(unbound):
return unbound.im_func
def create_bound_method(func, obj):
return types.MethodType(func, obj, obj.__class__)
class Iterator(object):
def next(self):
return type(self).__next__(self)
callable = callable
_add_doc(get_unbound_function,
"""Get the function out of a possibly unbound function""")
get_method_function = operator.attrgetter(_meth_func)
get_method_self = operator.attrgetter(_meth_self)
get_function_closure = operator.attrgetter(_func_closure)
get_function_code = operator.attrgetter(_func_code)
get_function_defaults = operator.attrgetter(_func_defaults)
get_function_globals = operator.attrgetter(_func_globals)
if PY3:
def iterkeys(d, **kw):
return iter(d.keys(**kw))
def itervalues(d, **kw):
return iter(d.values(**kw))
def iteritems(d, **kw):
return iter(d.items(**kw))
def iterlists(d, **kw):
return iter(d.lists(**kw))
viewkeys = operator.methodcaller("keys")
viewvalues = operator.methodcaller("values")
viewitems = operator.methodcaller("items")
else:
def iterkeys(d, **kw):
return iter(d.iterkeys(**kw))
def itervalues(d, **kw):
return iter(d.itervalues(**kw))
def iteritems(d, **kw):
return iter(d.iteritems(**kw))
def iterlists(d, **kw):
return iter(d.iterlists(**kw))
viewkeys = operator.methodcaller("viewkeys")
viewvalues = operator.methodcaller("viewvalues")
viewitems = operator.methodcaller("viewitems")
_add_doc(iterkeys, "Return an iterator over the keys of a dictionary.")
_add_doc(itervalues, "Return an iterator over the values of a dictionary.")
_add_doc(iteritems,
"Return an iterator over the (key, value) pairs of a dictionary.")
_add_doc(iterlists,
"Return an iterator over the (key, [values]) pairs of a dictionary.")
if PY3:
def b(s):
return s.encode("latin-1")
def u(s):
return s
unichr = chr
if sys.version_info[1] <= 1:
def int2byte(i):
return bytes((i,))
else:
# This is about 2x faster than the implementation above on 3.2+
int2byte = operator.methodcaller("to_bytes", 1, "big")
byte2int = operator.itemgetter(0)
indexbytes = operator.getitem
iterbytes = iter
import io
StringIO = io.StringIO
BytesIO = io.BytesIO
else:
def b(s):
return s
# Workaround for standalone backslash
def u(s):
return unicode(s.replace(r'\\', r'\\\\'), "unicode_escape")
unichr = unichr
int2byte = chr
def byte2int(bs):
return ord(bs[0])
def indexbytes(buf, i):
return ord(buf[i])
iterbytes = functools.partial(itertools.imap, ord)
import StringIO
StringIO = BytesIO = StringIO.StringIO
_add_doc(b, """Byte literal""")
_add_doc(u, """Text literal""")
if PY3:
exec_ = getattr(moves.builtins, "exec")
def reraise(tp, value, tb=None):
if value is None:
value = tp()
if value.__traceback__ is not tb:
raise value.with_traceback(tb)
raise value
else:
def exec_(_code_, _globs_=None, _locs_=None):
"""Execute code in a namespace."""
if _globs_ is None:
frame = sys._getframe(1)
_globs_ = frame.f_globals
if _locs_ is None:
_locs_ = frame.f_locals
del frame
elif _locs_ is None:
_locs_ = _globs_
exec ("""exec _code_ in _globs_, _locs_""")
exec_("""def reraise(tp, value, tb=None):
raise tp, value, tb
""")
if sys.version_info > (3, 2):
exec_("""def raise_from(value, from_value):
raise value from from_value
""")
else:
def raise_from(value, from_value):
raise value
print_ = getattr(moves.builtins, "print", None)
if print_ is None:
def print_(*args, **kwargs):
"""The new-style print function for Python 2.4 and 2.5."""
fp = kwargs.pop("file", sys.stdout)
if fp is None:
return
def write(data):
if not isinstance(data, basestring):
data = str(data)
# If the file has an encoding, encode unicode with it.
if (isinstance(fp, file) and
isinstance(data, unicode) and
fp.encoding is not None):
errors = getattr(fp, "errors", None)
if errors is None:
errors = "strict"
data = data.encode(fp.encoding, errors)
fp.write(data)
want_unicode = False
sep = kwargs.pop("sep", None)
if sep is not None:
if isinstance(sep, unicode):
want_unicode = True
elif not isinstance(sep, str):
raise TypeError("sep must be None or a string")
end = kwargs.pop("end", None)
if end is not None:
if isinstance(end, unicode):
want_unicode = True
elif not isinstance(end, str):
raise TypeError("end must be None or a string")
if kwargs:
raise TypeError("invalid keyword arguments to print()")
if not want_unicode:
for arg in args:
if isinstance(arg, unicode):
want_unicode = True
break
if want_unicode:
newline = unicode("\n")
space = unicode(" ")
else:
newline = "\n"
space = " "
if sep is None:
sep = space
if end is None:
end = newline
for i, arg in enumerate(args):
if i:
write(sep)
write(arg)
write(end)
_add_doc(reraise, """Reraise an exception.""")
if sys.version_info[0:2] < (3, 4):
def wraps(wrapped, assigned=functools.WRAPPER_ASSIGNMENTS,
updated=functools.WRAPPER_UPDATES):
def wrapper(f):
f = functools.wraps(wrapped)(f)
f.__wrapped__ = wrapped
return f
return wrapper
else:
wraps = functools.wraps
def with_metaclass(meta, *bases):
"""Create a base class with a metaclass."""
# This requires a bit of explanation: the basic idea is to make a dummy
# metaclass for one level of class instantiation that replaces itself with
# the actual metaclass.
class metaclass(meta):
def __new__(cls, name, this_bases, d):
return meta(name, bases, d)
return type.__new__(metaclass, 'temporary_class', (), {})
def add_metaclass(metaclass):
"""Class decorator for creating a class with a metaclass."""
def wrapper(cls):
orig_vars = cls.__dict__.copy()
slots = orig_vars.get('__slots__')
if slots is not None:
if isinstance(slots, str):
slots = [slots]
for slots_var in slots:
orig_vars.pop(slots_var)
orig_vars.pop('__dict__', None)
orig_vars.pop('__weakref__', None)
return metaclass(cls.__name__, cls.__bases__, orig_vars)
return wrapper
# Complete the moves implementation.
# This code is at the end of this module to speed up module loading.
# Turn this module into a package.
__path__ = [] # required for PEP 302 and PEP 451
__package__ = __name__ # see PEP 366 @ReservedAssignment
if globals().get("__spec__") is not None:
__spec__.submodule_search_locations = [] # PEP 451 @UndefinedVariable
# Remove other six meta path importers, since they cause problems. This can
# happen if six is removed from sys.modules and then reloaded. (Setuptools does
# this for some reason.)
if sys.meta_path:
for i, importer in enumerate(sys.meta_path):
# Here's some real nastiness: Another "instance" of the six module might
# be floating around. Therefore, we can't use isinstance() to check for
# the six meta path importer, since the other six instance will have
# inserted an importer with different class.
if (type(importer).__name__ == "_SixMetaPathImporter" and
importer.name == __name__):
del sys.meta_path[i]
break
del i, importer
# Finally, add the importer to the meta path import hook.
sys.meta_path.append(_importer)
python-casacore-2.2.1/casacore/tables/ 0000775 0000000 0000000 00000000000 13203113151 0017633 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/tables/__init__.py 0000775 0000000 0000000 00000005430 13203113151 0021751 0 ustar 00root root 0000000 0000000 # __init__.py: Top level .py file for python table interface
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: __init__.py,v 1.6 2006/11/06 01:54:21 gvandiep Exp $
"""Python interface to the Casacore tables module.
A `casacore table <../../casacore/doc/html/group__Tables__module.html>`_
is similar to a relational data base table with the extension
that table cells can contain n-dimensional arrays.
It has a rich SQL-like query language
(`TaQL <../../doc/199.html>`_).
A table consists of numbered rows and named columns. A column can hold
scalar values or arrays of any dimensionality and shape. Furthermore the
table and each column can hold a set of keywords (e.g. to define the units).
It is nestable, thus the value of a keyword can be a keyword set in itself.
The `tables` module consists of a few classes:
:class:`table`
main module to open, create, access, and query tables
:class:`tablecolumn`
access the contents of a column in an easier way
:class:`tablerow`
access the contents of table rows or parts of it
:class:`tableiter`
iterate through a table based on the contents of one or more columns
:class:`tableindex`
build and use an index on one or more table columns
submodule `tableutil <#table-utility-functions>`_
table utility functions (e.g. to create a table description)
submodule `msutil <#measurementset-utility-functions>`_
MeasuementSet utility functions (e.g. to concat MSs)
"""
from .msutil import *
from .table import table
from .table import default_ms
from .table import default_ms_subtable
from .table import tablecommand
from .table import taql
from .tablecolumn import tablecolumn
from .tableindex import tableindex
from .tableiter import tableiter
from .tablerow import tablerow
from .tableutil import *
python-casacore-2.2.1/casacore/tables/msutil.py 0000664 0000000 0000000 00000042306 13203113151 0021527 0 ustar 00root root 0000000 0000000 # msutil.py: Utility MeasurementSet functions
# Copyright (C) 2011
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
import numpy as np
from casacore import six
from casacore.tables.table import table, taql, _required_ms_desc
from casacore.tables.tableutil import makescacoldesc, makearrcoldesc, \
makecoldesc, maketabdesc
def required_ms_desc(table=None):
"""
Obtain the required table description for a given table.
If "" or "MAIN", the description for a MeasurementSet is returned.
Otherwise, a the description for a MeasurementSet subtable is returned.
"""
# Default to MAIN table
if table is None:
table = ""
return _required_ms_desc(table)
def addImagingColumns(msname, ack=True):
""" Add the columns to an MS needed for the casa imager.
It adds the columns MODEL_DATA, CORRECTED_DATA, and IMAGING_WEIGHT.
It also sets the CHANNEL_SELECTION keyword needed for the older casa
imagers.
A column is not added if already existing.
"""
# numpy is needed
import numpy as np
# Open the MS
t = table(msname, readonly=False, ack=False)
cnames = t.colnames()
# Get the description of the DATA column.
try:
cdesc = t.getcoldesc('DATA')
except:
raise ValueError('Column DATA does not exist')
# Determine if the DATA storage specification is tiled.
hasTiled = False
try:
dminfo = t.getdminfo("DATA")
if dminfo['TYPE'][:5] == 'Tiled':
hasTiled = True
except:
hasTiled = False
# Use TiledShapeStMan if needed.
if not hasTiled:
dminfo = {'TYPE': 'TiledShapeStMan',
'SPEC': {'DEFAULTTILESHAPE': [4, 32, 128]}}
# Add the columns(if not existing). Use the description of the DATA column.
if 'MODEL_DATA' in cnames:
six.print_("Column MODEL_DATA not added; it already exists")
else:
dminfo['NAME'] = 'modeldata'
cdesc['comment'] = 'The model data column'
t.addcols(maketabdesc(makecoldesc('MODEL_DATA', cdesc)), dminfo)
if ack:
six.print_("added column MODEL_DATA")
if 'CORRECTED_DATA' in cnames:
six.print_("Column CORRECTED_DATA not added; it already exists")
else:
dminfo['NAME'] = 'correcteddata'
cdesc['comment'] = 'The corrected data column'
t.addcols(maketabdesc(makecoldesc('CORRECTED_DATA', cdesc)), dminfo)
if ack:
six.print_("'added column CORRECTED_DATA")
if 'IMAGING_WEIGHT' in cnames:
six.print_("Column IMAGING_WEIGHT not added; it already exists")
else:
# Add IMAGING_WEIGHT which is 1-dim and has type float.
# It needs a shape, otherwise the CASA imager complains.
shp = []
if 'shape' in cdesc:
shp = cdesc['shape']
if len(shp) > 0:
shp = [shp[0]] # use nchan from shape
else:
shp = [t.getcell('DATA', 0).shape[0]] # use nchan from actual data
cd = makearrcoldesc('IMAGING_WEIGHT', 0, ndim=1, shape=shp,
valuetype='float')
dminfo = {'TYPE': 'TiledShapeStMan',
'SPEC': {'DEFAULTTILESHAPE': [32, 128]}}
dminfo['NAME'] = 'imagingweight'
t.addcols(maketabdesc(cd), dminfo)
if ack:
six.print_("added column IMAGING_WEIGHT")
# Add or overwrite keyword CHANNEL_SELECTION.
if 'CHANNEL_SELECTION' in t.colkeywordnames('MODEL_DATA'):
t.removecolkeyword('MODEL_DATA', 'CHANNEL_SELECTION')
# Define the CHANNEL_SELECTION keyword containing the channels of
# all spectral windows.
tspw = table(t.getkeyword('SPECTRAL_WINDOW'), ack=False)
nchans = tspw.getcol('NUM_CHAN')
chans = [[0, nch] for nch in nchans]
t.putcolkeyword('MODEL_DATA', 'CHANNEL_SELECTION', np.int32(chans))
if ack:
six.print_("defined keyword CHANNEL_SELECTION in column MODEL_DATA")
# Flush the table to make sure it is written.
t.flush()
def removeImagingColumns(msname):
# Open the MS
t = table(msname, readonly=False, ack=False)
# Remove if the column exists.
cnames = t.colnames()
removeNames = []
for col in ['MODEL_DATA', 'CORRECTED_DATA', 'IMAGING_WEIGHT']:
if col in cnames:
removeNames.append(col)
if len(removeNames) > 0:
t.removecols(removeNames)
t.flush()
def addDerivedMSCal(msname):
""" Add the derived columns like HA to an MS or CalTable.
It adds the columns HA, HA1, HA2, PA1, PA2, LAST, LAST1, LAST2, AZEL1,
AZEL2, and UVW_J2000.
They are all bound to the DerivedMSCal virtual data manager.
It fails if one of the columns already exists.
"""
# Open the MS
t = table(msname, readonly=False, ack=False)
colnames = t.colnames()
# Check that the columns needed by DerivedMSCal are present.
# Note that ANTENNA2 and FEED2 are not required.
for col in ["TIME", "ANTENNA1", "FIELD_ID", "FEED1"]:
if col not in colnames:
raise ValueError("Columns " + colnames +
" should be present in table " + msname)
scols1 = ['HA', 'HA1', 'HA2', 'PA1', 'PA2']
scols2 = ['LAST', 'LAST1', 'LAST2']
acols1 = ['AZEL1', 'AZEL2']
acols2 = ['UVW_J2000']
descs = []
# Define the columns and their units.
for col in scols1:
descs.append(makescacoldesc(col, 0.,
keywords={"QuantumUnits": ["rad"]}))
for col in scols2:
descs.append(makescacoldesc(col, 0.,
keywords={"QuantumUnits": ["d"]}))
for col in acols1:
descs.append(makearrcoldesc(col, 0.,
keywords={"QuantumUnits": ["rad", "rad"]}))
for col in acols2:
descs.append(makearrcoldesc(col, 0.,
keywords={"QuantumUnits": ["m", "m", "m"],
"MEASINFO": {"Ref": "J2000",
"type": "uvw"}}))
# Add all columns using DerivedMSCal as data manager.
dminfo = {"TYPE": "DerivedMSCal", "NAME": "", "SPEC": {}}
t.addcols(maketabdesc(descs), dminfo)
# Flush the table to make sure it is written.
t.flush()
def removeDerivedMSCal(msname):
""" Remove the derived columns like HA from an MS or CalTable.
It removes the columns using the data manager DerivedMSCal.
Such columns are HA, HA1, HA2, PA1, PA2, LAST, LAST1, LAST2, AZEL1,
AZEL2, and UVW_J2000.
It fails if one of the columns already exists.
"""
# Open the MS
t = table(msname, readonly=False, ack=False)
# Remove the columns stored as DerivedMSCal.
dmi = t.getdminfo()
for x in dmi.values():
if x['TYPE'] == 'DerivedMSCal':
t.removecols(x['COLUMNS'])
t.flush()
def msconcat(names, newname, concatTime=False):
"""Virtually concatenate multiple MeasurementSets.
Multiple MeasurementSets are concatenated into a single MeasurementSet.
The concatenation is done in an entirely or almost entirely virtual way,
so hardly any data are copied. It makes the command very fast and hardly
any extra disk space is needed.
The MSs can be concatenated in time or frequency (spectral windows).
If concatenated in time, no indices need to be updated and the
concatenation is done in a single step.
If spectral windows are concatenated, the data-description-ids and
spectral-window-ids in the resulting MS and its subtables are updated
to make them unique.
The spectral concatenation is done in two steps and results in two MSs:
1. The input MSs are virtually concatenated resulting in the
MeasurementSet `_CONCAT`.
2. The MeasurementSet is created. It references all columns
in `_CONCAT` with the exception of the DATA_DESC_ID column.
This column is copied and updated to make the ids correct.
Furthermore the MS contains a copy of all subtables (with the exception
of SORTED_TABLE), where the DATA_DESCRIPTION and SPECTRAL_WINDOW
subtables are the concatenation of those subtables in the input MSs.
The ids in the resulting subtables are updated.
The FEED, FREQ_OFFSET, SOURCE, and SYSCAL subtables also have a
SPECTRAL_WINDOW_ID column. Currently these subtables are not concatenated
nor are their ids updated.
`names`
A sequence containing the names of the MeasurementSets to concatenate.
`newname`
The name of the resulting MeasurementSet. A MeasurementSet with this
name followed by `_CONCAT` will also be created (and must be kept).
`concatTime`
False means that the spectral windows ids will be adjusted as explained
above.
"""
if len(names) == 0:
raise ValueError('No input MSs given')
# Concatenation in time is straightforward.
if concatTime:
t = table(names[0])
if 'SYSCAL' in t.fieldnames():
tn = table(names, concatsubtables='SYSCAL')
else:
tn = table(names)
t.close()
tn.rename(newname)
return
# First concatenate the given tables as another table.
# The SPECTRAL_WINDOW and DATA_DESCRIPTION subtables are concatenated
# and changed later.
# Those subtables cannot be concatenated here, because the deep copy of
# them fails due to the rename of the main table.
tn = table(names)
tdesc = tn.getdesc()
tn.rename(newname + '_CONCAT')
tn.flush()
# Now create a table where all columns forward to the concatenated table,
# but create a stored column for the data description id, because it has
# to be changed.
# The new column is filled at the end.
tnew = table(newname, tdesc, nrow=tn.nrows(), dminfo={
'1': {'TYPE': 'ForwardColumnEngine',
'NAME': 'ForwardData',
'COLUMNS': tn.colnames(),
'SPEC': {'FORWARDTABLE': tn.name()}}})
# Remove the DATA_DESC_ID column and recreate it in a stored way.
tnew.removecols('DATA_DESC_ID')
tnew.addcols(makecoldesc('DATA_DESC_ID', tdesc['DATA_DESC_ID']),
dminfo={'TYPE': 'IncrementalStMan',
'NAME': 'DDID',
'SPEC': {}})
# Copy the table keywords.
keywords = tn.getkeywords()
tnew.putkeywords(keywords)
# Copy all column keywords.
for col in tn.colnames():
tnew.putcolkeywords(col, tn.getcolkeywords(col))
# Make a deep copy of all subtables (except SORTED_TABLE).
for key in keywords:
if key != 'SORTED_TABLE':
val = keywords[key]
if isinstance(val, str):
tsub = table(val, ack=False)
tsubn = tsub.copy(newname + '/' + key, deep=True)
tnew.putkeyword(key, tsubn)
tnew.flush()
# Now we have to take care that the subbands are numbered correctly.
# The DATA_DESCRIPTION and SPECTRAL_WINDOW subtables are concatenated.
# The ddid in the main table and spwid in DD subtable have to be updated.
tnewdd = table(tnew.getkeyword('DATA_DESCRIPTION'),
readonly=False, ack=False)
tnewspw = table(tnew.getkeyword('SPECTRAL_WINDOW'),
readonly=False, ack=False)
nrdd = 0
nrspw = 0
nrmain = 0
for name in names:
t = table(name, ack=False)
tdd = table(t.getkeyword('DATA_DESCRIPTION'), ack=False)
tspw = table(t.getkeyword('SPECTRAL_WINDOW'), ack=False)
# The first table already has its subtable copied.
# Append the subtables of the other ones.
if nrdd > 0:
tnewdd.addrows(tdd.nrows())
for i in range(tdd.nrows()):
tnewdd[nrdd + i] = tdd[i] # copy row i
tnewspw.addrows(tspw.nrows())
for i in range(tspw.nrows()):
tnewspw[nrspw + i] = tspw[i]
tnewdd.putcol('SPECTRAL_WINDOW_ID',
tdd.getcol('SPECTRAL_WINDOW_ID') + nrspw,
nrdd, tdd.nrows())
tnew.putcol('DATA_DESC_ID',
t.getcol('DATA_DESC_ID') + nrdd,
nrmain, t.nrows())
nrdd += tdd.nrows()
nrspw += tspw.nrows()
nrmain += t.nrows()
# Overwrite keyword CHANNEL_SELECTION.
if 'MODEL_DATA' in tnew.colnames():
if 'CHANNEL_SELECTION' in tnew.colkeywordnames('MODEL_DATA'):
tnew.removecolkeyword('MODEL_DATA', 'CHANNEL_SELECTION')
# Define the CHANNEL_SELECTION keyword containing the channels of
# all spectral windows.
tspw = table(tnew.getkeyword('SPECTRAL_WINDOW'), ack=False)
nchans = tspw.getcol('NUM_CHAN')
chans = [[0, nch] for nch in nchans]
tnew.putcolkeyword('MODEL_DATA', 'CHANNEL_SELECTION',
np.int32(chans))
# Future work:
# If SOURCE subtables have to concatenated, the FIELD and DOPPLER
# have to be dealt with as well.
# The FEED table can be concatenated; the FEED_ID can stay the same,
# but spwid has to be updated.
# The FREQ_OFFSET table is stand-alone, thus can simply be concatenated
# and have spwid updated.
# The SYSCAL table can be very large, so it might be better to virtually
# concatenate it instead of making a copy (just like the main table).
# Flush the table and subtables.
tnew.flush(True)
def msregularize(msname, newname):
""" Regularize an MS
The output MS will be such that it has the same number of baselines
for each time stamp. Where needed fully flagged rows are added.
Possibly missing rows are written into a separate MS -add.
It is concatenated with the original MS and sorted in order of TIME,
DATADESC_ID, ANTENNA1,ANTENNA2 to form a new regular MS. Note that
the new MS references the input MS (it does not copy the data).
It means that changes made in the new MS are also made in the input MS.
If no rows were missing, the new MS is still created referencing the
input MS.
"""
# Find out all baselines.
t = table(msname)
t1 = t.sort('unique ANTENNA1,ANTENNA2')
nadded = 0
# Now iterate in time,band over the MS.
for tsub in t.iter(['TIME', 'DATA_DESC_ID']):
nmissing = t1.nrows() - tsub.nrows()
if nmissing < 0:
raise ValueError("A time/band chunk has too many rows")
if nmissing > 0:
# Rows needs to be added for the missing baselines.
ant1 = str(t1.getcol('ANTENNA1')).replace(' ', ',')
ant2 = str(t1.getcol('ANTENNA2')).replace(' ', ',')
ant1 = tsub.getcol('ANTENNA1')
ant2 = tsub.getcol('ANTENNA2')
t2 = taql('select from $t1 where !any(ANTENNA1 == $ant1 &&' +
' ANTENNA2 == $ant2)')
six.print_(nmissing, t1.nrows(), tsub.nrows(), t2.nrows())
if t2.nrows() != nmissing:
raise ValueError("A time/band chunk behaves strangely")
# If nothing added yet, create a new table.
# (which has to be reopened for read/write).
# Otherwise append to that new table.
if nadded == 0:
tnew = t2.copy(newname + "_add", deep=True)
tnew = table(newname + "_add", readonly=False)
else:
t2.copyrows(tnew)
# Set the correct time and band in the new rows.
tnew.putcell('TIME',
range(nadded, nadded + nmissing),
tsub.getcell('TIME', 0))
tnew.putcell('DATA_DESC_ID',
range(nadded, nadded + nmissing),
tsub.getcell('DATA_DESC_ID', 0))
nadded += nmissing
# Combine the existing table and new table.
if nadded > 0:
# First initialize data and flags in the added rows.
taql('update $tnew set DATA=0+0i')
taql('update $tnew set FLAG=True')
tcomb = table([t, tnew])
tcomb.rename(newname + '_adds')
tcombs = tcomb.sort('TIME,DATA_DESC_ID,ANTENNA1,ANTENNA2')
else:
tcombs = t.query(offset=0)
tcombs.rename(newname)
six.print_(newname, 'has been created; it references the original MS')
if nadded > 0:
six.print_(' and', newname + '_adds', 'containing', nadded, 'new rows')
else:
six.print_(' no rows needed to be added')
python-casacore-2.2.1/casacore/tables/table.py 0000775 0000000 0000000 00000227627 13203113151 0021317 0 ustar 00root root 0000000 0000000 # table.py: Python table functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: table.py,v 1.13 2006/11/10 01:18:53 gvandiep Exp $
"""Access to Casacore tables.
The :class:`table` class is the main class to access a table. Its constructor
can open or create a table.
Several utility functions exist. Important ones are:
:func:`taql` (or its synonym `tablecommand`)
executes a TaQL query command and returns a :class:`table` object.
:func:`tablefromascii`
creates a table from an ASCII file and returns a :class:`table` object.
"""
# Make interface to class TableProxy available.
from ._tables import (Table,
_default_ms,
_default_ms_subtable,
_required_ms_desc)
from .tablehelper import (_add_prefix, _remove_prefix, _do_remove_prefix, _format_row)
from casacore import six
def default_ms(name, tabdesc=None, dminfo=None):
"""
Creates a default Measurement Set called name. Any Table Description
elements in tabdesc will overwrite the corresponding element in a default
Measurement Set Table Description (columns, hypercolumns and keywords).
In practice, you probably want to specify columns such as DATA, MODEL_DATA
and CORRECTED_DATA (and their associated keywords and hypercolumns) in tabdesc
"""
# Default to empty dictionaries
if tabdesc is None:
tabdesc = {}
if dminfo is None:
dminfo = {}
# Wrap the Table object
return table(_default_ms(name, tabdesc, dminfo), _oper=3)
def default_ms_subtable(subtable, name=None, tabdesc=None, dminfo=None):
"""
Creates a default Measurement Set subtable. Any Table Description
elements in tabdesc will overwrite the corresponding element in a default
Measurement Set Table Description (columns, hypercolumns and keywords).
if name is given, it will be treated as a path that the table should
be created in. Set to subtable if None
if subtable is "" or "MAIN" a standard MeasurementSet with subtables will
be created.
"""
if name is None:
name = subtable
# Default to empty dictionaries
if tabdesc is None:
tabdesc = {}
if dminfo is None:
dminfo = {}
# Wrap the Table object
return table(_default_ms_subtable(subtable, name, tabdesc, dminfo), _oper=3)
# Execute a TaQL command on a table.
def taql(command, style='Python', tables=[], globals={}, locals={}):
"""Execute a TaQL command and return a table object.
A `TaQL <../../doc/199.html>`_
command is an SQL-like command to do a selection of rows and/or
columns in a table.
The default style used in a TaQL command is python, which means 0-based
indexing, C-ordered arrays, and non-inclusive end in ranges.
It is possible to use python variables directly in the command using
`$var` where `var` is the name of the variable to use. For example::
t = table('3c343.MS')
value = 5.1
t1 = taql('select from $t where COL > $value')
In this example the table `$t` is replaced by a sequence number (like `$1`)
and `$value` by its value 5.1.
The table object of `t` will be appended to a copy of the `tables` argument
such that the sequence number inserted matches the table object in the list.
The more advanced user can already use `$n` in the query string and
supply the associated table object in the `tables` argument
(where `n` represents the (n-1)th `tables` element).
The :func:`query` command makes use of this feature.
The arguments `globals` and `locals` can be used to pass in a dict
containing the possible variables used in the TaQL command. They can
be obtained with the python functions locals() and globals().
If `locals` is empty, the local variables in the calling function will
be used, so normally one does not need to use these arguments.
"""
# Substitute possible tables given as $name.
cmd = command
# Copy the tables argument and make sure it is a list
tabs = []
for tab in tables:
tabs += [tab]
try:
import casacore.util
if len(locals) == 0:
# local variables in caller are 3 levels up from getlocals
locals = casacore.util.getlocals(3)
cmd = casacore.util.substitute(cmd, [(table, '', tabs)], globals, locals)
except:
pass
if style:
cmd = 'using style ' + style + ' ' + cmd
tab = table(cmd, tabs, _oper=2)
result = tab._getcalcresult()
# If result is empty, it was a normal TaQL command resulting in a table.
# Otherwise it is a record containing calc values.
if len(result) == 0:
return tab
return result['values']
# alias
tablecommand = taql
class table(Table):
"""The Python interface to Casacore tables.
One can open or create tables, get or put data in them, make selections,
get meta information (like storage managers used), etc.
It is possible to lock/unlock a table for concurrent access.
A table consists of numbered rows and named columns. A column can hold
scalar values or arrays of any dimensionality and shape. Furthermore the
table and each column can hold a possibly nested keyword set
(e.g. to define the units).
The classes :class:`tablecolumn`, :class:`tableiter`, :class:`tableindex`,
and :class:`tablerow` tablecolumn can be used for easier access to tables.
Module tableutil contains some useful utility functions, for instance
:func:`tablefromascii` to create a table from an ASCII file.
Several functions accept or return arrays for which numpy arrays are used.
One dimensional arrays can also be passed as sequences (e.g., a list).
A scalar value can also be passed to functions expecting an array and
results in a 1-dim array of length 1.
Scalar arguments can be passed as normal python scalars, but also as
numpy scalars (which have a special data type).
If needed and if possible, data type conversion is done automatically.
A `table` object contains a :class:`tablerow` object which contains
all columns. Similar to `tablerow` the `table` object can be indexed
in the standard python way to get (or put) values in one or more rows.
For example::
t = table('~/3c343.MS')
print t[0]
The `table` class supports the context manager idiom (__enter__ and __exit__).
When used in a `with` statement, the table will be flushed and closed
automatically, which is handy when writing a table.
For example::
with table('my.ms') as t:
t.putcell ('SPECTRAL_WINDOW_ID', 0, 0)
Usually a table is kept on disk, but it can also reside in memory.
Furthermore, results of sort and selection are kept as so-called
reference tables which are kept in memory (but can be made persistent).
`table('tablename', tabledesc)`
creates a new table using the given table description which can be
obtained from an existing table (using table.getdesc) or created
(using tableutil.maketabdesc).
If `memorytable=True`, the table is created in memory.
`table('tablename', readonly=False)`
| opens an existing table for read/write. Default is for readonly.
| In general it is a bad idea to open a subtable using a path like
'my.ms/ANTENNA', because it will fail if 'my.ms' is a selection instead
of a plain table. Therefore a double colon can be used like
'my.ms::ANTENNA' making the table system handle it in a correct way.
`table(['table1','table2',...])`
opens a virtual table as the concatenation of the given tables.
The tables have to have the same columns.
`table([tableobject1,tableobject2,...])`
opens a virtual table as the concatenation of the given table objects.
The tables have to have the same columns.
The following arguments can be used.
`tablename`
| name of the table to be opened or created. As in most UNIX shells
a table name can contain environment variables and a tilde with
optional user name.
| This argument can be a sequence of names in which case they are
opened as a (virtual) concatenation of tables.
| It can also be a sequence of table objects to be concatenated.
`tabledesc`
description of table to be created. If given as a dict (which should
be created by :func:`maketabdesc`, the table is created. Otherwise
it should exist and is opened.
`nrow`
initial number of rows in table to be created (default 0).
`readonly = True/False`
tell if a table has to be opened readonly (default) or read/write.
`lockoptions`
see below
`ack=False`
prohibit the printing of a message telling if the table was
opened or created successfully.
`dminfo`
a dict (as returned by :func:`getdminfo`) giving specific data manager
info for one or more columns. In this way expert users can
tell how data are stored (or use virtual data managers).
`endian`
| endianness of the data in table to be created
| 'little' = as little endian
| 'big' = as big endian
| 'local' = use the endianness of the machine being used
| 'aipsrc' = use a defined in an .aipsrc file (defaults to local)
`memorytable=True`
create table in memory instead of on disk.
`concatsubtables`
if tables are concatenated, it should contain a sequence of the
names of subtables to be concatenated as well (default none).
Non-mentioned subtables are considered to be identical in each table,
so only the subtable of the first table is used as subtable for the
concatenated table.
Locking/unlocking to share a table in a concurrent environment is
controlled by the lockoptions argument.
`auto`
let the system take care of locking. It locks when needed,
but unlocking is usually not done automatically.
`autonoread`
as `auto`, but no read locking is needed. This must be
used with care, because it means that reading can be done
while the table object is not synchronized with the table
file (as is normally done when a lock is acquired).
The method :func:`resync` can be used to explicitly synchronize
the table object with the file.
`user`
the user takes care by explicit calls to lock and unlock
`usernoread`
as `user` and the no readlocking behaviour of `autonoread`.
`permanent`
use a permanent write lock; the constructor fails if the
lock cannot be acquired because the table is already
in use in another process
`permanentwait`
as above, but wait until the lock is acquired.
`default`
this is the default option.
If the given table is already open, the locking option
in use is not changed. Otherwise it reverts to `auto`.
If auto locking is used, it is possible to give a record containing
the fields option, interval, and/or maxwait (see :func:`lockoptions`
for their meaning). In this way advanced users can have full control
over the locking process. In practice this is hardly ever needed.
For example::
t = table('3c343.ms') # open table readonly
t = table('3c343.ms', readonly=False) # open table read/write
t1= table('new.tab', t.getdesc()) # create table
t = table([t1,t2,t3,t4]) # concatenate 4 tables
"""
def __init__(self, tablename, tabledesc=False, nrow=0, readonly=True,
lockoptions='default', ack=True, dminfo={}, endian='aipsrc',
memorytable=False, concatsubtables=[],
_columnnames=[], _datatypes=[],
_oper=0, _delete=False):
"""Open or create a table."""
if _oper == 1:
# This is the readascii constructor.
tabname = _remove_prefix(tablename)
Table.__init__(self, tabname, tabledesc, nrow, readonly,
lockoptions, ack, dminfo, endian, memorytable,
_columnnames, _datatypes)
elif _oper == 2:
# This is the query or calc constructor.
Table.__init__(self, tablename, tabledesc)
if len(self._getcalcresult()) != 0:
# Do not make row object for a calc result
return
elif _oper == 3:
# This is the constructor taking a Table (used by copy).
Table.__init__(self, tablename)
else:
# This is the constructor for a normal table open.
# It can be done in several forms:
# - open single existing table (PlainTable)
# - open multiple existing tables (ConcatTable)
# - create a new table (PlainTable or MemoryTable)
# - concatenate open tables (ConcatTable)
tabname = _remove_prefix(tablename)
lockopt = lockoptions
if isinstance(lockoptions, str):
lockopt = {'option': lockoptions}
if isinstance(tabledesc, dict):
# Create a new table.
memtype = 'plain'
if (memorytable):
memtype = 'memory'
Table.__init__(self, tabname, lockopt, endian,
memtype, nrow, tabledesc, dminfo)
if ack:
six.print_('Successful creation of', lockopt['option'] + '-locked table', tabname + ':',
self.ncols(), 'columns,', self.nrows(), 'rows')
else:
# Deal with existing tables.
if not tabname:
raise ValueError("No tables or names given")
# Open an existing table
opt = 1
typstr = 'readonly'
if not readonly:
typstr = 'read/write'
opt = 5
if _delete:
opt = 6
if isinstance(tabname, str):
Table.__init__(self, tabname, lockopt, opt)
if ack:
six.print_('Successful', typstr, 'open of', lockopt['option'] + '-locked table', tabname + ':',
self.ncols(), 'columns,', self.nrows(), 'rows')
elif isinstance(tabname[0], str):
# Concatenate and open named tables.
Table.__init__(self, tabname, concatsubtables, lockopt, opt)
if ack:
six.print_('Successful', typstr, 'open of', lockopt['option'] + '-locked concatenated tables',
tabname, ':', self.ncols(), 'columns,', self.nrows(), 'rows')
else:
# Concatenate already open tables.
Table.__init__(self, tabname, concatsubtables, 0, 0, 0)
if ack:
six.print_('Successful virtual concatenation of', len(tabname), 'tables:', self.ncols(),
'columns,', self.nrows(), 'rows')
# Create a row object for this table.
self._makerow()
def __enter__(self):
"""Function to enter a with block."""
return self
def __exit__(self, type, value, traceback):
"""Function to exit a with block which closes the table object."""
self.close()
def _makerow(self):
"""Internal method to make its tablerow object."""
from .tablerow import _tablerow
self._row = _tablerow(self, [])
def __str__(self):
"""Return the table name and the basic statistics"""
return _add_prefix(self.name()) + "\n%d rows" % self.nrows() + "\n" + \
"%d columns: " % len(self.colnames()) + " ".join(self.colnames())
def __len__(self):
"""Return the number of rows in the table."""
return self._nrows()
def __getattr__(self, name):
"""Get the tablecolumn object or keyword value.
| A tablecolumn object is returned if it names a column.
| The value of a keyword is returned if it names a keyword.
If the keyword is a subtable, it opens the table and returns a
table object.
| The values of all keywords is returned if name equals _ or keys.
An AttributeError is raised if the name is column nor keyword.
For example::
print t.DATA[i] # print row i of column DATA
print t.MS_VERSION # print the MS version
print t.keys # print values of all keywords
subtab = t.FEED # open the FEED subtable
"""
# First try if it is a column.
try:
return self.col(name)
except:
pass
# Now try if it is a keyword.
try:
val = self.getkeyword(name)
# See if the keyword represents a subtable and try to open it.
if val != _do_remove_prefix(val):
try:
return table(val, ack=False)
except:
pass
return val
except:
pass
# _ or keys means all keywords.
if name == '_' or name == 'keys':
return self.getkeywords()
# Unknown name.
raise AttributeError("table has no attribute/column/keyword " + name)
def __getitem__(self, key):
"""Get the values from one or more rows."""
return self._row._getitem(key, self.nrows())
def __setitem__(self, key, value):
"""Put value into one or more rows."""
self._row._setitem(key, value, self.nrows())
def col(self, columnname):
"""Return a tablecolumn object for the given column.
If multiple operations need to be done on a column, a :class:`tablecolumn`
object is somewhat easier to use than the table object because the
column name does not have to be repeated each time.
It is also possible to use a column name as an attribute, It is an
easier way to get a column object.
For example::
tc = t.col('DATA')
tc[0:10] # get first 10 rows of column DATA
t.DATA[0:10] # does the same in an easier way
"""
from .tablecolumn import tablecolumn
return tablecolumn(self, columnname)
def row(self, columnnames=[], exclude=False):
"""Return a tablerow object which includes (or excludes) the given columns.
:class:`tablerow` makes it possible to get/put values in one or
more rows.
"""
from .tablerow import tablerow
return tablerow(self, columnnames, exclude)
def iter(self, columnnames, order='', sort=True):
"""Return a tableiter object.
:class:`tableiter` lets one iterate over a table by returning in each
iteration step a reference table containing equal values for the given
columns.
By default a sort is done on the given columns to get the correct
iteration order.
`order`
| 'ascending' is iterate in ascending order (is the default).
| 'descending' is iterate in descending order.
`sort=False`
do not sort (because table is already in correct order).
For example, iterate by time through a measurementset table::
t = table('3c343.MS')
for ts in t.iter('TIME'):
print ts.nrows()
"""
from .tableiter import tableiter
return tableiter(self, columnnames, order, sort)
def index(self, columnnames, sort=True):
"""Return a tableindex object.
:class:`tableindex` lets one get the row numbers of the rows holding
given values for the columns for which the index is created.
It uses an in-memory index on which a binary search is done.
By default the table is sorted on the given columns to get the correct
index order.
For example::
t = table('3c343.MS')
tinx = t.index('ANTENNA1')
print tinx.rownumbers(0) # print rownrs containing ANTENNA1=0
"""
from .tableindex import tableindex
return tableindex(self, columnnames, sort)
def flush(self, recursive=False):
"""Flush the table to disk.
Until a flush or unlock is performed, the results of operations might
not be stored on disk yet.
| If `recursive=True`, all subtables are flushed as well.
"""
self._flush(recursive)
def resync(self):
"""Resync the table object with the file contents.
Usually concurrent access is handled by acquiring read and write locks.
However, a table can be opened without the need for read locking using
lock option `usernoread` or `autonoread`. in that case
synchronization of the table object and actual file contents can be
done manually using this method.
"""
self._resync()
def close(self):
"""Flush and close the table which invalidates the table object."""
self._row = 0
self._close()
def done(self):
"""Flush and close the table which invalidates the table object."""
self.close()
def toascii(self, asciifile, headerfile='', columnnames=(), sep=' ',
precision=(), usebrackets=True):
"""Write the table in ASCII format.
It is approximately the inverse of the from-ASCII-contructor.
`asciifile`
The name of the resulting ASCII file.
`headerfile`
The name of an optional file containing the header info. If not
given or if equal to argument `asciifile`, the headers are written
at the beginning of the ASCII file.
`columnnames`
The names of the columns to be written. If not given or if the first
name is empty, all columns are written.
`sep`
The separator to be used between values. Only the first character
of a string is used. If not given or mepty, a blank is used.
`precision`
For each column the precision can be given. It is only used for
columns containing floating point numbers. A value <=0 means using
the default which is 9 for single and 18 for double precision.
`usebrackets`
If True, arrays and records are written enclosed in [].
Multi-dimensional arrays have [] per dimension. In this way variable
shaped array can be read back correctly. However, it is not supported
by :func:`tablefromascii`.
If False, records are not written and arrays are written linearly
with the shape defined in the header as supported byI
:func:`tablefromascii`.
Note that columns containing records or variable shaped arrays are
ignored, because they cannot be written to ASCII. It is told which
columns are ignored.
For example::
t = table('3c343.MS')
t1 = t.query('ANTENNA1 != ANTENNA2') # do row selection
t1.toascii ('3c343.txt') # write selection as ASCII
"""
msg = self._toascii(asciifile, headerfile, columnnames, sep,
precision, usebrackets)
if len(msg) > 0:
six.print_(msg)
def rename(self, newtablename):
"""Rename the table.
It renames the table and, if needed, adjusts the names of its subtables.
"""
self._rename(newtablename)
def copy(self, newtablename, deep=False, valuecopy=False, dminfo={},
endian='aipsrc', memorytable=False, copynorows=False):
"""Copy the table and return a table object for the copy.
It copies all data in the columns and keywords.
Besides the table, all its subtables are copied too.
By default a shallow copy is made (usually by copying files).
It means that the copy of a reference table is also a reference table.
Use `deep=True` to make a deep copy which turns a reference table
into a normal table.
`deep=True`
a deep copy of a reference table is made.
`valuecopy=True`
values are copied, which reorganizes normal tables and removes wasted
space. It implies `deep=True`. It is slower than a normal copy.
`dminfo`
gives the option to specify data managers to change the way columns
are stored. This is a dict as returned by method :func:`getdminfo`.
`endian`
specifies the endianness of the new table when a deep copy is made:
| 'little' = as little endian
| 'big' = as big endian
| 'local' = use the endianness of the machine being used
| 'aipsrc' = use as defined in an .aipsrc file (defaults to local)
`memorytable=True`
do not copy to disk, but to a table kept in memory.
`copynorows=True`
only copy the column layout and keywords, but no data.
For example::
t = table('3c343.MS')
t1 = t.query('ANTENNA1 != ANTENNA2') # do row selection
t2 = t1.copy ('3c343.sel', True) # make deep copy
t2 = t.copy ('new.tab', True, True) # reorganize storage
"""
t = self._copy(newtablename, memorytable, deep, valuecopy,
endian, dminfo, copynorows)
# copy returns a Table object, so turn that into table.
return table(t, _oper=3)
def copyrows(self, outtable, startrowin=0, startrowout=-1, nrow=-1):
"""Copy the contents of rows from this table to outtable.
The contents of the columns with matching names are copied.
The other arguments can be used to specify where to start copying.
By default the entire input table is appended to the output table.
Rows are added to the output table if needed.
`startrowin`
Row where to start in the input table.
`startrowout`
Row where to start in the output table,
| -1 means write at the end of the output table.
`nrow`
Number of rows to copy
| -1 means from startrowin till the end of the input table
The following example appends row to the table itself, thus doubles
the number of rows::
t:=table('test.ms',readonly=F)
t.copyrows(t)
"""
self._copyrows(outtable, startrowin, startrowout, nrow)
def iswritable(self):
"""Return if the table is writable."""
return self._iswritable()
def endianformat(self):
"""Return the endian format ('little' or 'big') in which the table is written."""
return self._endianformat()
def lock(self, write=True, nattempts=0):
"""Acquire a read or write lock on a table.
`write=False` means a read lock, otherwise a write lock.
| nattempts defines the nr of attempts (one attempt per second) to do
before giving up. The default 0 means unlimited.
An exception is thrown if no lock could be acquired.
If the table has already been locked appropriately, nothing will be
done. Thus locks do NOT nest.
"""
self._lock(write, nattempts)
def unlock(self):
"""Unlock the table.
Flush the table data and release a read or write lock on the table
acquired by lock().
Nothing will be done if the table is not locked.
"""
self._unlock()
def haslock(self, write=True):
"""Test if the table is read or write locked."""
return self._haslock(write)
def lockoptions(self):
"""Return the lockoptions.
They are returned as a dict with fields:
'option'
the locking mode (user, usernoread, auto, autonoread, permanent,
permanentwait).
`interval`
In case of AutoLocking the inspection interval defines how often
the table system checks if another process needs a lock on the table.
`maxwait`
the maximum time to wait when acquiring a lock in AutoLocking mode.
"""
return self._lockoptions()
def datachanged(self):
"""Tell if data in the table have changed since the last time called."""
return self._datachanged()
def ismultiused(self, checksubtables=False):
"""Tell if the table is used in other processes.
`checksubtables=True` means it will also check it for subtables.
"""
return self._ismultiused(checksubtables)
def name(self):
"""Return the table name."""
return self._name()
def partnames(self, recursive=False):
"""Return the names of the tables this table consists of.
A table can be a reference to another table (e.g. for a selection) or
a concatenation of other tables. This function returns the names of
such table parts.
For a plain table it simply returns the name of that table.
In its turn a table part can be a reference or concatenated table.
`recursive=True` means that it follows table parts until the end.
"""
return self._partnames(recursive)
def info(self):
"""Return the table info (table type, subtype, and readme lines)."""
return self._info()
def putinfo(self, value):
"""Put the table info.
The table info is a dict containing the fields:
"""
self._putinfo(value)
def addreadmeline(self, value):
"""Add a readme line to the table info."""
self._addreadmeline(value)
def setmaxcachesize(self, columnname, nbytes):
"""Set the maximum cache size for the data manager used by the column.
It can sometimes be useful to limit the size of the cache used by
a column stored with the tiled storage manager.
This method requires some more knowledge about the table system
and is not meant for the casual user.
"""
self._setmaxcachesize(columnname, nbytes)
def rownumbers(self, table=None):
"""Return a list containing the row numbers of this table.
This method can be useful after a selection or a sort.
It returns the row numbers of the rows in this table with respect
to the given table. If no table is given, the original table is used.
For example::
t = table('W53.MS')
t1 = t.selectrows([1,3,5,7,9]) # select a few rows
t1.rownumbers(t)
# [1 3 5 7 9]
t2 = t1.selectrows([2,5]) # select rows from the selection
t2.rownumbers(t1)
# [2 5] # rownrs of t2 in table t1
t2.rownumbers(t)
# [3 9] # rownrs of t2 in t
t2.rownumbers()
# [3 9]
The last statements show that the method returns the row numbers
referring to the given table. Table t2 contains rows 2 and 5 in table t1,
which are rows 3 and 9 in table t.
"""
if table is None:
return self._rownumbers(Table())
return self._rownumbers(table)
def colnames(self):
"""Get the names of all columns in the table."""
return self._colnames()
def isscalarcol(self, columnname):
"""Tell if the column contains scalar values."""
return self._isscalarcol(columnname)
def isvarcol(self, columnname):
"""Tell if the column holds variable shaped arrays."""
desc = self.getcoldesc(columnname)
return 'ndim' in desc and 'shape' not in desc
def coldatatype(self, columnname):
"""Get the data type of a column.
It returns a string which can have the values:
``boolean integer float double complex dcomplex string record``
"""
return self._coldatatype(columnname)
def colarraytype(self, columnname):
"""Get the array type of a column holding arrays.
It tells if an array is fixed or variable shaped and if it is stored
directly or indirectly. This is done by means of a string like
``Indirect, variable sized arrays``
"""
return self._colarraytype(columnname)
def ncols(self):
"""Return the number of columns in the table."""
return self._ncols()
def nrows(self):
"""Return the number of rows in the table."""
return self._nrows()
def addrows(self, nrows=1):
"""Add one or more rows to the table."""
self._addrows(nrows)
def removerows(self, rownrs):
"""Remove the given rows from the table.
The row numbers can be given in a sequence in any order. The rows will
be removed from the end of the table towards the beginning. This is
needed because the removal of a row decrements the row number of higher
rows.
Thus::
t.removerow ([10,20])
is different from::
t.removerow(10)
t.removerow(20)
because in the latter row 20 was row 21 before the removal of row 10.
Some storage managers (in particular the tiled ones) do not allow
removal of rows, so the operation may fail.
Rows can always be removed from a reference table. It does NOT remove
the rows from the referenced table.
"""
self._removerows(rownrs)
def getcolshapestring(self, columnname,
startrow=0, nrow=-1, rowincr=1):
"""Get the shapes of all cells in the column in string format.
It returns the shape in a string like [10,20,30].
If the column contains fixed shape arrays, a single shape is returned.
Otherwise a list of shape strings is returned.
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
"""
return self._getcolshapestring(columnname,
startrow, nrow, rowincr,
True); # reverse axes
def iscelldefined(self, columnname, rownr):
"""Tell if a column cell contains a value.
Columns containing variable shaped arrays can be empty. For these cases
this method returns True. Doing :func:`getcell` on an empty cell results
in an exception.
Note that an empty cell is not the same as an empty array. A cell can
contain an empty array (of any dimensionality) as a value.
Also note that a cell in a column containing scalars or fixed shaped
arrays cannot be empty.
"""
return self._iscelldefined(columnname, rownr)
def getcell(self, columnname, rownr):
"""Get data from a column cell.
Get the contents of a cell which can be returned as a scalar value,
a numpy array, or a dict depending on the contents of the cell.
"""
return self._getcell(columnname, rownr)
def getcellnp(self, columnname, rownr, nparray):
"""Get data from a column cell into the given numpy array .
Get the contents of a cell containing an array into the
given numpy array. The numpy array has to be C-contiguous
with a shape matching the shape of the column cell.
Data type coercion will be done as needed.
"""
if not nparray.flags.c_contiguous or nparray.size == 0:
raise ValueError("Argument 'nparray' has to be a contiguous numpy array")
return self._getcellvh(columnname, rownr, nparray)
def getcellslice(self, columnname, rownr, blc, trc, inc=[]):
"""Get a slice from a column cell holding an array.
The columnname and (0-relative) rownr indicate the table cell.
The slice to get is defined by the blc, trc, and optional inc arguments.
(blc = bottom-left corner, trc=top-right corner, inc=stride). Not all
axes have to be filled in for blc, trc, and inc. Missing axes default
to begin, end, and 1. A negative blc or trc defaults to begin or end.
Note that trc is inclusive (unlike python indexing).
"""
return self._getcellslice(columnname, rownr,
blc, trc, inc)
def getcellslicenp(self, columnname, nparray, rownr, blc, trc, inc=[]):
"""Get a slice from a column cell into the given numpy array.
The columnname and (0-relative) rownr indicate the table cell.
The numpy array has to be C-contiguous with a shape matching the
shape of the slice. Data type coercion will be done as needed.
The slice to get is defined by the blc, trc, and optional inc arguments.
(blc = bottom-left corner, trc=top-right corner, inc=stride). Not all
axes have to be filled in for blc, trc, and inc. Missing axes default
to begin, end, and 1. A negative blc or trc defaults to begin or end.
Note that trc is inclusive (unlike python indexing).
"""
if not nparray.flags.c_contiguous or nparray.size == 0:
raise ValueError("Argument 'nparray' has to be a contiguous numpy array")
return self._getcellslicevh(columnname, rownr,
blc, trc, inc, nparray)
def getcol(self, columnname, startrow=0, nrow=-1, rowincr=1):
"""Get the contents of a column or part of it.
It is returned as a numpy array.
If the column contains arrays, they should all have the same shape.
An exception is thrown if they differ in shape. In that case the
method :func:`getvarcol` should be used instead.
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
"""
# try: # trial code to read using a vector of rownrs
# nr = len(startrow)
# if nrow < 0:
# nrow = nr
# if nrow == 0:
# return numpy.array()
# for inx in range(nrow):
# i = inx*
# except:
return self._getcol(columnname, startrow, nrow, rowincr)
def getcolnp(self, columnname, nparray, startrow=0, nrow=-1, rowincr=1):
"""Get the contents of a column or part of it into the given numpy array.
The numpy array has to be C-contiguous with a shape matching the
shape of the column (part). Data type coercion will be done as needed.
If the column contains arrays, they should all have the same shape.
An exception is thrown if they differ in shape. In that case the
method :func:`getvarcol` should be used instead.
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
"""
if (not nparray.flags.c_contiguous) or nparray.size == 0:
raise ValueError("Argument 'nparray' has to be a contiguous numpy array")
return self._getcolvh(columnname, startrow, nrow, rowincr, nparray)
def getvarcol(self, columnname, startrow=0, nrow=-1, rowincr=1):
"""Get the contents of a column or part of it.
It is similar to :func:`getcol`, but the result is returned as a dict of
numpy arrays.
It can deal with a column containing variable shaped arrays.
"""
return self._getvarcol(columnname, startrow, nrow, rowincr)
def getcolslice(self, columnname, blc, trc, inc=[],
startrow=0, nrow=-1, rowincr=1):
"""Get a slice from a table column holding arrays.
The slice in each array is given by blc, trc, and inc (as in getcellslice).
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
It returns a numpy array where the first axis is formed by the column
cells. The other axes are the array axes.
"""
return self._getcolslice(columnname, blc, trc, inc,
startrow, nrow, rowincr)
def getcolslicenp(self, columnname, nparray, blc, trc, inc=[],
startrow=0, nrow=-1, rowincr=1):
"""Get a slice from a table column into the given numpy array.
The numpy array has to be C-contiguous with a shape matching the
shape of the column (slice). Data type coercion will be done as needed.
The slice in each array is given by blc, trc, and inc (as in getcellslice).
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
It returns a numpy array where the first axis is formed by the column
cells. The other axes are the array axes.
"""
if not nparray.flags.c_contiguous or nparray.size == 0:
raise ValueError("Argument 'nparray' has to be a contiguous numpy array")
return self._getcolslicevh(columnname, blc, trc, inc,
startrow, nrow, rowincr, nparray)
def putcell(self, columnname, rownr, value):
"""Put a value into one or more table cells.
The columnname and (0-relative) rownrs indicate the table cells.
rownr can be a single row number or a sequence of row numbers.
If multiple rownrs are given, the given value is put in all those rows.
The given value has to be convertible to the data type of the column.
If the column contains scalar values, the given value must be a scalar.
The value for a column holding arrays can be given as:
- a scalar resulting in a 1-dim array of 1 element
- a sequence (list, tuple) resulting in a 1-dim array
- a numpy array of any dimensionality
Note that the arrays in a column may have a fixed dimensionality or
shape. In that case the dimensionality or shape of the array to put
has to conform.
"""
self._putcell(columnname, rownr, value)
def putcellslice(self, columnname, rownr, value, blc, trc, inc=[]):
"""Put into a slice of a table cell holding an array.
The columnname and (0-relative) rownr indicate the table cell.
Unlike putcell only a single row can be given.
The slice to put is defined by the blc, trc, and optional inc arguments.
(blc = bottom-left corner, trc=top-right corner, inc=stride). Not all
axes have to be filled in for blc, trc, and inc. Missing axes default
to begin, end, and 1. A negative blc or trc defaults to begin or end.
Note that trc is inclusive (unlike python indexing).
As in putcell the array can be given by a scalar, sequence, or numpy array.
The shape of the array to put has to match the slice shape.
"""
self._putcellslice(columnname, rownr, value,
blc, trc, inc)
def putcol(self, columnname, value, startrow=0, nrow=-1, rowincr=1):
"""Put an entire column or part of it.
If the column contains scalar values, the given value should be a 1-dim
array. Otherwise it is a numpy array where the first axis is formed by
the column cells.
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
"""
self._putcol(columnname, startrow, nrow, rowincr, value)
def putvarcol(self, columnname, value, startrow=0, nrow=-1, rowincr=1):
"""Put an entire column or part of it.
It is similar to putcol, but the shapes of the arrays in the column
can vary. The value has to be a dict of numpy arrays.
The column can be sliced by giving a start row (default 0), number of
rows (default all), and row stride (default 1).
"""
self._putvarcol(columnname, startrow, nrow, rowincr, value)
def putcolslice(self, columnname, value, blc, trc, inc=[],
startrow=0, nrow=-1, rowincr=1):
"""Put into a slice in a table column holding arrays.
Its arguments are the same as for getcolslice and putcellslice.
"""
self._putcolslice(columnname, value, blc, trc, inc,
startrow, nrow, rowincr)
def addcols(self, desc, dminfo={}, addtoparent=True):
"""Add one or more columns.
Columns can always be added to a normal table.
They can also be added to a reference table and optionally to its
parent table.
`desc`
contains a description of the column(s) to be added. It can be given
in three ways:
- a dict created by :func:`maketabdesc`. In this way multiple
columns can be added.
- a dict created by :func:`makescacoldesc`, :func:`makearrcoldesc`,
or :func:`makecoldesc`. In this way a single column can be added.
- a dict created by :func:`getcoldesc`. The key 'name' containing
the column name has to be defined in such a dict.
`dminfo`
can be used to provide detailed data manager info to tell how the
column(s) have to be stored. The dminfo of an existing column can be
obtained using method :func:`getdminfo`.
`addtoparent`
defines if the column should also be added to the parent table in
case the current table is a reference table (result of selection).
If True, it will be added to the parent if it does not exist yet.
For example, add a column using the same data manager type as another
column::
coldmi = t.getdminfo('colarrtsm') # get dminfo of existing column
coldmi["NAME"] = 'tsm2' # give it a unique name
t.addcols (maketabdesc(makearrcoldesc("colarrtsm2",0., ndim=2)),
coldmi)
"""
tdesc = desc
# Create a tabdesc if only a coldesc is given.
if 'name' in desc:
import casacore.tables.tableutil as pt
if len(desc) == 2 and 'desc' in desc:
# Given as output from makecoldesc
tdesc = pt.maketabdesc(desc)
elif 'valueType' in desc:
# Given as output of getcoldesc (with a name field added)
cd = pt.makecoldesc(desc['name'], desc)
tdesc = pt.maketabdesc(cd)
self._addcols(tdesc, dminfo, addtoparent)
self._makerow()
def renamecol(self, oldname, newname):
"""Rename a single table column.
Renaming a column in a reference table does NOT rename the column in
the referenced table.
"""
self._renamecol(oldname, newname)
self._makerow()
def removecols(self, columnnames):
"""Remove one or more columns.
Note that some storage managers (in particular the tiled ones) do not
allow the removal of one of its columns. In that case all its columns
have to be removed together.
Columns can always be removed from a reference table. It does NOT remove
the columns from the referenced table.
"""
self._removecols(columnnames)
self._makerow()
def keywordnames(self):
"""Get the names of all table keywords."""
return self._getfieldnames('', '', -1)
def colkeywordnames(self, columnname):
"""Get the names of all keywords of a column."""
return self._getfieldnames(columnname, '', -1)
def fieldnames(self, keyword=''):
"""Get the names of the fields in a table keyword value.
The value of a keyword can be a struct (python dict). This method
returns the names of the fields in that struct.
Each field in a struct can be a struct in itself. Names of fields in a
sub-struct can be obtained by giving a keyword name consisting of
multiple parts separated by dots (e.g. 'key1.sub1.sub2').
If an empty keyword name is given (which is the default), all table
keyword names are shown and its behaviour is the same as :func:`keywordnames`.
Instead of a keyword name an index can be given which returns the names
of the struct value of the i-th keyword.
"""
if isinstance(keyword, str):
return self._getfieldnames('', keyword, -1)
else:
return self._getfieldnames('', '', keyword)
def colfieldnames(self, columnname, keyword=''):
"""Get the names of the fields in a column keyword value.
The value of a keyword can be a struct (python dict). This method
returns the names of the fields in that struct.
Each field in a struct can be a struct in itself. Names of fields in a
sub-struct can be obtained by giving a keyword name consisting of
multiple parts separated by dots (e.g. 'key1.sub1.sub2').
If an empty keyword name is given (which is the default), all keyword
names of the column are shown and its behaviour is the same as
:func:`colkeywordnames`.
Instead of a keyword name an index can be given which returns the names
of the struct value of the i-th keyword.
"""
if isinstance(keyword, str):
return self._getfieldnames(columnname, keyword, -1)
else:
return self._getfieldnames(columnname, '', keyword)
def getkeyword(self, keyword):
"""Get the value of a table keyword.
The value of a keyword can be a:
- scalar which is returned as a normal python scalar.
- an array which is returned as a numpy array.
- a reference to a table which is returned as a string containing its
name prefixed by 'Table :'. It can be opened using the normal table
constructor which will remove the prefix.
- a struct which is returned as a dict. A struct is fully nestable,
thus each field in the struct can have one of the values described here.
Similar to method :func:`fieldnames` a keyword name can be given consisting
of multiple parts separated by dots. This represents nested structs,
thus gives the value of a field in a struct (in a struct, etc.).
Instead of a keyword name an index can be given which returns the value
of the i-th keyword.
"""
if isinstance(keyword, str):
return self._getkeyword('', keyword, -1)
else:
return self._getkeyword('', '', keyword)
def getcolkeyword(self, columnname, keyword):
"""Get the value of a column keyword.
It is similar to :func:`getkeyword`.
"""
if isinstance(keyword, str):
return self._getkeyword(columnname, keyword, -1)
else:
return self._getkeyword(columnname, '', keyword)
def getkeywords(self):
"""Get the value of all table keywords.
It is returned as a dict. See :func:`getkeyword` for the possible
value types.
"""
return self._getkeywords('')
def getcolkeywords(self, columnname):
"""Get the value of all keywords of a column.
It is returned as a dict. See :func:`getkeyword` for the possible
value types.
"""
return self._getkeywords(columnname)
def getsubtables(self):
"""Get the names of all subtables."""
keyset = self.getkeywords()
names = []
for key, value in keyset.items():
if isinstance(value, str) and value.find('Table: ') == 0:
names.append(_do_remove_prefix(value))
return names
def putkeyword(self, keyword, value, makesubrecord=False):
"""Put the value of a table keyword.
The value of a keyword can be a:
- scalar which can be given a normal python scalar or numpy scalar.
- an array which can be given as a numpy array. A 1-dimensional array
can also be given as a sequence (tuple or list).
- a reference to a table which can be given as a table object or as a
string containing its name prefixed by 'Table :'.
- a struct which can be given as a dict. A struct is fully nestable,
thus each field in the dict can be one of the values described here.
The only exception is that a table value can only be given by the string.
If the keyword already exists, the type of the new value should match
the existing one (e.g. a scalar cannot be replaced by an array).
Similar to method :func:`getkeyword` a keyword name can be given
consisting of multiple parts separated by dots. This represents nested
structs, thus puts the value into a field in a struct (in a struct,
etc.).
If `makesubrecord=True` structs will be created for the keyword name
parts that do not exist.
Instead of a keyword name an index can be given which returns the value
of the i-th keyword.
"""
val = value
if isinstance(val, table):
val = _add_prefix(val.name())
if isinstance(keyword, str):
return self._putkeyword('', keyword, -1, makesubrecord, val)
else:
return self._putkeyword('', '', keyword, makesubrecord, val)
def putcolkeyword(self, columnname, keyword, value, makesubrecord=False):
"""Put the value of a column keyword.
It is similar to :func:`putkeyword`.
"""
val = value
if isinstance(val, table):
val = _add_prefix(val.name())
if isinstance(keyword, str):
return self._putkeyword(columnname, keyword, -1,
makesubrecord, val)
else:
return self._putkeyword(columnname, '', keyword,
makesubrecord, val)
def putkeywords(self, value):
"""Put the value of multiple table keywords.
The value has to be a dict, so each field in the dict is a keyword.
It puts all keywords similar to :func:`putkeyword`.
"""
return self._putkeywords('', value)
def putcolkeywords(self, columnname, value):
"""Put the value of multiple keywords in a column.
The value has to be a dict, so each field in the dict is a keyword.
It puts all keywords similar to :func:`putkeyword`.
"""
return self._putkeywords(columnname, value)
def removekeyword(self, keyword):
"""Remove a table keyword.
Similar to :func:`getkeyword` the name can consist of multiple parts.
In that case a field in a struct will be removed.
Instead of a keyword name an index can be given which removes
the i-th keyword.
"""
if isinstance(keyword, str):
self._removekeyword('', keyword, -1)
else:
self._removekeyword('', '', keyword)
def removecolkeyword(self, columnname, keyword):
"""Remove a column keyword.
It is similar to :func:`removekeyword`.
"""
if isinstance(keyword, str):
self._removekeyword(columnname, keyword, -1)
else:
self._removekeyword(columnname, '', keyword)
def getdesc(self, actual=True):
"""Get the table description.
By default it returns the actual description (thus telling the
actual array shapes and data managers used).
`actual=False` means that the original description as made by
:func:`maketabdesc` is returned.
"""
tabledesc = self._getdesc(actual, True)
# Strip out 0 length "HCcoordnames" and "HCidnames"
# as these aren't valid. (See tabledefinehypercolumn)
hcdefs = tabledesc.get('_define_hypercolumn_', {})
for c, hcdef in hcdefs.iteritems():
if "HCcoordnames" in hcdef and len(hcdef["HCcoordnames"]) == 0:
del hcdef["HCcoordnames"]
if "HCidnames" in hcdef and len(hcdef["HCidnames"]) == 0:
del hcdef["HCidnames"]
return tabledesc
def getcoldesc(self, columnname, actual=True):
"""Get the description of a column.
By default it returns the actual description (thus telling the
actual array shapes and data managers used).
`actual=False` means that the original description as made by
:func:`makescacoldesc` or :func:`makearrcoldesc` is returned.
"""
return self._getcoldesc(columnname, actual, True)
def coldesc(self, columnname, actual=True):
"""Make the description of a column.
Make the description object of the given column like :func:`makecoldesc`
is doing with the description given by :func:`getcoldesc`.
"""
import casacore.tables.tableutil as pt
return pt.makecoldesc(columnname, self.getcoldesc(columnname, actual))
def getdminfo(self, columnname=None):
"""Get data manager info.
Each column in a table is stored using a data manager. A storage manager
is a data manager storing the physically in a file. A virtual column
engine is a data manager that does not store data but calculates it on
the fly (e.g. scaling floats to short to reduce storage).
By default this method returns a dict telling the data managers used.
Each field in the dict is a dict containing:
- NAME telling the (unique) name of the data manager
- TYPE telling the type of data manager (e.g. TiledShapeStMan)
- SEQNR telling the sequence number of the data manager
(is ''i'' in table.f for storage managers)
- SPEC is a dict holding the data manager specification
- COLUMNS is a list giving the columns stored by this data manager
When giving a column name the data manager info of that particular
column is returned (without the COLUMNS field).
It can, for instance, be used when adding a column using :func:`addcols`
that should use the same data manager type as an existing column.
However, when doing that care should be taken to change the NAME because
each data manager name has to be unique.
"""
dminfo = self._getdminfo()
if columnname is None:
return dminfo
# Find the info for the given column
for fld in dminfo.values():
if columnname in fld["COLUMNS"]:
fldc = fld.copy()
del fldc['COLUMNS'] # remove COLUMNS field
return fldc
raise KeyError("Column " + columnname + " does not exist")
def getdmprop(self, name, bycolumn=True):
"""Get properties of a data manager.
Each column in a table is stored using a data manager. A storage manager
is a data manager storing the physically in a file. A virtual column
engine is a data manager that does not store data but calculates it on
the fly (e.g. scaling floats to short to reduce storage).
Some data managers have properties that can be changed on the fly
(e.g. cachesize for a tiled storage manager). The properties of
a given data manager are returned as a dict; function :func:`setdmprop`
can be used to change the properties. Note the properties are also part
of the data manager info returned by :func:`getdminfo`.
The data manager can be specified in two ways: by data manager name
or by the name of a column using the data manager. The argument
`bycolumn` defines which way is used (default is by column name).
"""
return self._getdmprop(name, bycolumn)
def setdmprop(self, name, properties, bycolumn=True):
"""Set properties of a data manager.
Properties (e.g. cachesize) of a data manager can be changed by
defining them appropriately in the properties argument (a dict).
Current values can be obtained using function :func:`getdmprop` which
also serves as a template. The dict can contain more fields; only
the fields with the names as returned by getdmprop are handled.
The data manager can be specified in two ways: by data manager name
or by the name of a column using the data manager. The argument
`bycolumn` defines which way is used (default is by column name).
"""
return self._setdmprop(name, properties, bycolumn)
def showstructure(self, dataman=True, column=True, subtable=False,
sort=False):
"""Show table structure in a formatted string.
The structure of this table and optionally its subtables is shown.
It shows the data manager info and column descriptions.
Optionally the columns are sorted in alphabetical order.
`dataman`
Show data manager info? If False, only column info is shown.
If True, data manager info and columns per data manager are shown.
`column`
Show column description per data manager? Only takes effect if
dataman=True.
`subtable`
Show the structure of all subtables (recursively).
The names of subtables are always shown.
'sort'
Sort the columns in alphabetical order?
"""
return self._showstructure(dataman, column, subtable, sort)
def summary(self, recurse=False):
"""Print a summary of the table.
It prints the number of columns and rows, column names, and table and
column keywords.
If `recurse=True` it also prints the summary of all subtables, i.e.
tables referenced by table keywords.
"""
six.print_('Table summary:', self.name())
six.print_('Shape:', self.ncols(), 'columns by', self.nrows(), 'rows')
six.print_('Info:', self.info())
tkeys = self.getkeywords()
if (len(tkeys) > 0):
six.print_('Table keywords:', tkeys)
columns = self.colnames()
if (len(columns) > 0):
six.print_('Columns:', columns)
for column in columns:
ckeys = self.getcolkeywords(column)
if (len(ckeys) > 0):
six.print_(column, 'keywords:', ckeys)
if (recurse):
for key, value in tkeys.items():
tabname = _remove_prefix(value)
six.print_('Summarizing subtable:', tabname)
lt = table(tabname)
if (not lt.summary(recurse)):
break
return True
def selectrows(self, rownrs):
"""Return a reference table containing the given rows."""
t = self._selectrows(rownrs, name='')
# selectrows returns a Table object, so turn that into table.
return table(t, _oper=3)
def query(self, query='', name='', sortlist='', columns='',
limit=0, offset=0, style='Python'):
"""Query the table and return the result as a reference table.
This method queries the table. It forms a
`TaQL <../../doc/199.html>`_
command from the given arguments and executes it using the
:func:`taql` function.
The result is returned in a so-called reference table which references
the selected columns and rows in the original table. Usually a reference
table is temporary, but it can be made persistent by giving it a name.
Note that a reference table is handled as any table, thus can be
queried again.
All arguments are optional, but at least one of `query`, `name`,
`sortlist`, and `columns` should be used.
See the `TaQL note <../../doc/199.html>`_ for the
detailed description of the the arguments representing the various
parts of a TaQL command.
`query`
The WHERE part of a TaQL command.
`name`
The name of the reference table if it is to be made persistent.
`sortlist`
The ORDERBY part of a TaQL command. It is a single string in which
commas have to be used to separate sort keys.
`columns`
The columns to be selected (projection in data base terms). It is a
single string in which commas have to be used to separate column
names. Apart from column names, expressions can be given as well.
`limit`
If > 0, maximum number of rows to be selected.
`offset`
If > 0, ignore the first N matches.
`style`
The TaQL syntax style to be used (defaults to Python).
"""
if not query and not sortlist and not columns and limit <= 0 and offset <= 0:
raise ValueError('No selection done (arguments query, sortlist, columns, limit, and offset are empty)')
command = 'select '
if columns:
command += columns
command += ' from $1'
if query:
command += ' where ' + query
if sortlist:
command += ' orderby ' + sortlist
if limit > 0:
command += ' limit %d' % limit
if offset > 0:
command += ' offset %d' % offset
if name:
command += ' giving ' + name
return tablecommand(command, style, [self])
def sort(self, sortlist, name='',
limit=0, offset=0, style='Python'):
"""Sort the table and return the result as a reference table.
This method sorts the table. It forms a
`TaQL <../../doc/199.html>`_
command from the given arguments and executes it using the
:func:`taql` function.
The result is returned in a so-called reference table which references
the columns and rows in the original table. Usually a reference
table is temporary, but it can be made persistent by giving it a name.
Note that a reference table is handled as any table, thus can be
queried again.
`sortlist`
The ORDERBY part of a TaQL command. It is a single string in which
commas have to be used to separate sort keys. A sort key can be the
name of a column, but it can be an expression as well.
`name`
The name of the reference table if it is to be made persistent.
`limit`
If > 0, maximum number of rows to be selected after the sort step.
It can, for instance, be used to select the N highest values.
`offset`
If > 0, ignore the first `offset` matches after the sort step.
`style`
The TaQL syntax style to be used (defaults to Python).
"""
command = 'select from $1 orderby ' + sortlist
if limit > 0:
command += ' limit %d' % limit
if offset > 0:
command += ' offset %d' % offset
if name:
command += ' giving ' + name
return tablecommand(command, style, [self])
def select(self, columns, name='', style='Python'):
"""Select columns and return the result as a reference table.
This method represents the SELECT part of a TaQL command using the
given columns (or column expressions). It forms a
`TaQL <../../doc/199.html>`_
command from the given arguments and executes it using the
:func:`taql` function.
The result is returned in a so-called reference table which references
the columns and rows in the original table. Usually a reference
table is temporary, but it can be made persistent by giving it a name.
Note that a reference table is handled as any table, thus can be
queried again.
`columns`
The columns to be selected (projection in data base terms). It is a
single string in which commas have to be used to separate column
names. Apart from column names, expressions can be given as well.
`name`
The name of the reference table if it is to be made persistent.
`style`
The TaQL syntax style to be used (defaults to Python).
"""
command = 'select ' + columns + ' from $1'
if name:
command += ' giving ' + name
return tablecommand(command, style, [self])
def calc(self, expr, style='Python'):
"""Do a TaQL calculation
The TaQL CALC command can be used to get the result of a calculation on
table data. It is, however, also possible to use it without table data.
For instance, to use it for converting units::
t = table('',{})
t.calc ('(1 \in)cm')
`expr`
The CALC expression
`style`
The TaQL syntax style to be used (defaults to Python).
"""
return tablecommand('calc from $1 calc ' + expr, style, [self])
def browse(self, wait=True, tempname="/tmp/seltable"):
""" Browse a table using casabrowser or a simple wxwidget based browser.
By default the casabrowser is used if it can be found (in your PATH).
Otherwise the wxwidget one is used if wx can be loaded.
The casabrowser can only browse tables that are persistent on disk.
This gives problems for tables resulting from a query because they are
held in memory only (unless an output table name was given).
To make browsing of such tables possible, the argument `tempname` can
be used to specify a table name that will be used to form a persistent
table that can be browsed. Note that such a table is very small as it
does not contain data, but only references to rows in the original
table.
The default for `tempname` is '/tmp/seltable'.
If needed, the table can be deleted using the :func:`tabledelete`
function.
If `wait=False`, the casabrowser is started in the background.
In that case the user should delete a possibly created copy of a
temporary table.
"""
import os
# Test if casabrowser can be found.
# On OS-X 'which' always returns 0, so use test on top of it.
if os.system('test -x `which casabrowser` > /dev/null 2>&1') == 0:
waitstr1 = ""
waitstr2 = "foreground ..."
if not wait:
waitstr1 = " &"
waitstr2 = "background ..."
if self.iswritable():
six.print_("Flushing data and starting casabrowser in the " +
waitstr2)
else:
six.print_("Starting casabrowser in the " + waitstr2)
self.flush()
self.unlock()
if os.system('test -e ' + self.name() + '/table.dat') == 0:
os.system('casabrowser ' + self.name() + waitstr1)
elif len(tempname) > 0:
six.print_(" making a persistent copy in table " + tempname)
self.copy(tempname)
os.system('casabrowser ' + tempname + waitstr1)
if wait:
from casacore.tables import tabledelete
six.print_(" finished browsing")
tabledelete(tempname)
else:
six.print_(" after browsing use tabledelete('" + tempname +
"') to delete the copy")
else:
six.print_("Cannot browse because the table is in memory only.")
six.print_("You can browse a (shallow) persistent copy of the table like: ")
six.print_(" t.browse(True, '/tmp/tab1')")
else:
try:
import wxPython
except ImportError:
six.print_('casabrowser nor wx available')
return
from wxPython.wx import wxPySimpleApp
import sys
app = wxPySimpleApp()
from wxtablebrowser import CasaTestFrame
frame = CasaTestFrame(None, sys.stdout, self)
frame.Show(True)
app.MainLoop()
def view(self, wait=True, tempname="/tmp/seltable"):
""" View a table using casaviewer, casabrowser, or wxwidget based browser.
The table is viewed depending on the type:
MeasurementSet
is viewed using casaviewer.
Image
is viewed using casaviewer.
other
are browsed using the :func:`browse` function.
If the casaviewer cannot be found, all tables are browsed.
The casaviewer can only display tables that are persistent on disk.
This gives problems for tables resulting from a query because they are
held in memory only (unless an output table name was given).
To make viewing of such tables possible, the argument `tempname` can
be used to specify a table name that will be used to form a persistent
table that can be browsed. Note that such a table is very small as it
does not contain data, but only references to rows in the original table.
The default for `tempname` is '/tmp/seltable'.
If needed, the table can be deleted using the :func:`tabledelete`
function.
If `wait=False`, the casaviewer is started in the background.
In that case the user should delete a possibly created copy of a
temporary table.
"""
import os
# Determine the table type.
# Test if casaviewer can be found.
# On OS-X 'which' always returns 0, so use test on top of it.
viewed = False
type = self.info()["type"]
if type == "Measurement Set" or type == "Image":
if os.system('test -x `which casaviewer` > /dev/null 2>&1') == 0:
waitstr1 = ""
waitstr2 = "foreground ..."
if not wait:
waitstr1 = " &"
waitstr2 = "background ..."
if self.iswritable():
six.print_("Flushing data and starting casaviewer in the " + waitstr2)
else:
six.print_("Starting casaviewer in the " + waitstr2)
self.flush()
self.unlock()
if os.system('test -e ' + self.name() + '/table.dat') == 0:
os.system('casaviewer ' + self.name() + waitstr1)
viewed = True
elif len(tempname) > 0:
six.print_(" making a persistent copy in table " + tempname)
self.copy(tempname)
os.system('casaviewer ' + tempname + waitstr1)
viewed = True
if wait:
from casacore.tables import tabledelete
six.print_(" finished viewing")
tabledelete(tempname)
else:
six.print_(" after viewing use tabledelete('" + tempname + "') to delete the copy")
else:
six.print_("Cannot browse because the table is in memory only.")
six.print_("You can browse a (shallow) persistent copy of the table like:")
six.print_(" t.view(True, '/tmp/tab1')")
# Could not view the table, so browse it.
if not viewed:
self.browse(wait, tempname)
def _repr_html_(self):
"""Give a nice representation of tables in notebooks."""
out="
\n"
# Print column names (not if they are all auto-generated)
if not(all([colname[:4]=="Col_" for colname in self.colnames()])):
out+="
"
for colname in self.colnames():
out+="
"+colname+"
"
out+="
"
cropped=False
rowcount=0
for row in self:
rowout=_format_row(row, self.colnames(),self)
rowcount+=1
out+=rowout
if "\n" in rowout: # Double space after multiline rows
out+="\n"
out+="\n"
if rowcount>=20:
cropped=True
break
if out[-2:]=="\n\n":
out=out[:-1]
out+="
"
if cropped:
out+="
("+str(self.nrows()-20)+" more rows)
\n"
return out
python-casacore-2.2.1/casacore/tables/tablecolumn.py 0000664 0000000 0000000 00000030546 13203113151 0022522 0 ustar 00root root 0000000 0000000 # tablecolumn.py: Python tablecolumn functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: tablecolumn.py,v 1.9 2007/08/28 07:22:18 gvandiep Exp $
from .table import table
from .tablehelper import _check_key_slice, _do_remove_prefix, _format_cell
class tablecolumn:
"""The Python interface to a column in a Casacore table.
The `tablecolumn` class is a convenience class to access data in a
table column. All functionality provided in this class is available in
:class:`table`, but `tablecolumn` is more convenient to use because the
column name does not have to be given over and over again.
For example::
t = table('3C343.MS')
tc = tablecolumn(t, 'DATA')
# tc = t.col('DATA') # another way to construct a tablecolumn
tc.getcell(0) # get data from cell 0
As can be seen in the example :func:`table.col` offers a slightly more
convenient way to create a `tablecolumn` object.
A `tablecolumn` can be indexed using Python's [] operator. Negative start,
end, and stride is possible. For example::
tc[0] # get cell 0
tc[:5] # get cell 0,1,2,3,4
tc[-5,-1,] # get last 4 cells
tc[-1,-5,-1] # get last 4 cells in reversed order
tc[1] = tr[0] # put value of cell 0 into cell 1
The `tablecolumn` class supports the context manager idiom (__enter__ and __exit__).
When used in a `with` statement, the table changes will be flushed
automatically, which is handy when writing to the table column.
For example::
with t.SPECTRAL_WINDOW_ID as tc:
tc.putcell (0, 0)
"""
def __init__(self, table, columnname):
if columnname not in table.colnames():
raise RuntimeError("Column " + columnname +
" does not exist in table " + table.name())
self._table = table
self._column = columnname
def __enter__(self):
"""Function to enter a with block."""
return self
def __exit__(self, type, value, traceback):
"""Function to exit a with block which flushes the table object."""
self._table.flush()
def name(self):
"""Get the name of the column."""
return self._column
def table(self):
"""Get the table object this column belongs to."""
return self._table
def isscalar(self):
"""Tell if the column contains scalar values."""
return self._table.isscalarcol(self._column)
def isvar(self):
"""Tell if the column holds variable shaped arrays."""
return self._table.isvarcol(self._column)
def datatype(self):
"""Get the data type of the column.
(see :func:`table.coldatatype`)"""
return self._table.coldatatype(self._column)
def arraytype(self):
"""Get the array type of a column holding arrays.
(see :func:`table.colarraytype`)"""
return self._table.colarraytype(self._column)
def nrows(self):
"""Get number of cells in the column."""
return self._table.nrows()
def getshapestring(self, startrow=1, nrow=-1, rowincr=1):
"""Get the shapes of all cells in the column in string format.
(see :func:`table.getcolshapestring`)"""
return self._table.getcolshapestring(self._column,
startrow, nrow, rowincr)
def iscelldefined(self, rownr):
"""Tell if a column cell contains a value.
(see :func:`table.iscelldefined`)"""
return self._table.iscelldefined(self._column, rownr)
def getcell(self, rownr):
"""Get data from a column cell.
(see :func:`table.getcell`)"""
return self._table.getcell(self._column, rownr)
def getcellslice(self, rownr, blc, trc, inc=[]):
"""Get a slice from a column cell holding an array.
(see :func:`table.getcellslice`)"""
return self._table.getcellslice(self._column, rownr, blc, trc, inc)
def getcol(self, startrow=0, nrow=-1, rowincr=1):
"""Get the contents of the column or part of it.
(see :func:`table.getcol`)"""
return self._table.getcol(self._column, startrow, nrow, rowincr)
def getvarcol(self, startrow=0, nrow=-1, rowincr=1):
"""Get the contents of the column or part of it.
(see :func:`table.getvarcol`)"""
return self._table.getvarcol(self._column, startrow, nrow, rowincr)
def getcolslice(self, blc, trc, inc=[], startrow=0, nrow=-1, rowincr=1):
"""Get a slice from a table column holding arrays.
(see :func:`table.getcolslice`)"""
return self._table.getcolslice(self._column, blc, trc, inc, startrow, nrow, rowincr)
def putcell(self, rownr, value):
"""Put a value into one or more table cells.
(see :func:`table.putcell`)"""
return self._table.putcell(self._column, rownr, value)
def putcellslice(self, rownr, value, blc, trc, inc=[]):
"""Put into a slice of a table cell holding an array.
(see :func:`table.putcellslice`)"""
return self._table.putcellslice(self._column, rownr, value, blc, trc, inc)
def putcol(self, value, startrow=0, nrow=-1, rowincr=1):
"""Put an entire column or part of it.
(see :func:`table.putcol`)"""
return self._table.putcol(self._column, value, startrow, nrow, rowincr)
def putvarcol(self, value, startrow=0, nrow=-1, rowincr=1):
"""Put an entire column or part of it.
(see :func:`table.putvarcol`)"""
return self._table.putvarcol(self._column, value, startrow, nrow, rowincr)
def putcolslice(self, value, blc, trc, inc=[], startrow=0, nrow=-1, rowincr=1):
"""Put into a slice in a table column holding arrays.
(see :func:`table.putcolslice`)"""
return self._table.putcolslice(self._column, value, blc, trc, inc, startrow, nrow, rowincr)
def keywordnames(self):
"""Get the names of all keywords of the column."""
return self._table.colkeywordnames(self._column)
def fieldnames(self, keyword=''):
"""Get the names of the fields in a column keyword value.
(see :func:`table.colfieldnames`)"""
return self._table.colfieldnames(self._column, keyword)
def getkeyword(self, keyword):
"""Get the value of a column keyword.
(see :func:`table.getcolkeyword`)"""
return self._table.getcolkeyword(self._column, keyword)
def getkeywords(self):
"""Get the value of all keywords of the column.
(see :func:`table.getcolkeywords`)"""
return self._table.getcolkeywords(self._column)
def putkeyword(self, keyword, value, makesubrecord=False):
"""Put the value of a column keyword.
(see :func:`table.putcolkeyword`)"""
return self._table.putcolkeyword(self._column, keyword, value, makesubrecord)
def putkeywords(self, value):
"""Put the value of multiple table keywords.
(see :func:`table.putcolkeywords`)"""
return self._table.putcolkeywords(self._column, value)
def removekeyword(self, keyword):
"""Remove a column keyword.
(see :func:`table.removecolkeyword`)"""
return self._table.removecolkeyword(self._column, keyword)
def getdesc(self):
"""Get the description of the column.
(see :func:`table.getcoldesc`)"""
return self._table.getcoldesc(self._column)
def getdminfo(self):
"""Get data manager info of the column.
(see :func:`table.getdminfo`)"""
return self._table.getdminfo(self._column)
def iter(self, order='', sort=True):
"""Return a :class:`tableiter` object on this column."""
from casacore.tables import tableiter
return tableiter(self._table, [self._column], order, sort)
def index(self, sort=True):
"""Return a :class:`tableindex` object on this column."""
from casacore.tables import tableindex
return tableindex(self._table, [self._column], sort)
def __len__(self):
return self._table.nrows()
def __getattr__(self, name):
"""Get the keyword value.
| The value of a column keyword is returned if it names a keyword.
If the keyword is a subtable, it opens the table and returns a
table object.
| The values of all column keywords is returned if name equals _ or keys.
An AttributeError is raised if the name is not a keyword.
For example::
print tc.MEASINFO # print the column's measure info
print tc._ # print all column keywords
"""
# Try if it is a keyword.
try:
val = self.getkeyword(name)
# See if the keyword represents a subtable and try to open it.
if val != _do_remove_prefix(val):
try:
return table(val, ack=False)
except:
pass
return val
except:
pass
# _ or keys means all keywords.
if name in ('_', 'keys'):
return self.getkeywords()
# Unknown name.
raise AttributeError("table has no attribute/keyword " + name)
def __getitem__(self, key):
"""Get the values from one or more rows."""
sei = _check_key_slice(key, self._table.nrows(), 'tablecolumn')
if len(sei) == 1:
# A single row.
return self.getcell(sei[0])
# Handle row by row and store values in a list.
result = []
rownr = sei[0]
inx = 0
while inx < sei[1]:
result.append(self.getcell(rownr))
rownr += sei[2]
inx += 1
return result
def __setitem__(self, key, value):
sei = _check_key_slice(key, self._table.nrows(), 'tablecolumn')
if len(sei) == 1:
# A single row.
return self.putcell(sei[0], value)
# Handle row by row.
rownr = sei[0]
inx = 0
if not (isinstance(value, list) or isinstance(value, tuple)):
# The same value is put in all rows.
while inx < sei[1]:
self.putcell(rownr, value)
rownr += sei[2]
inx += 1
else:
# Each row has its own value.
if len(value) != sei[1]:
raise RuntimeError(
"tablecolumn slice length differs from value length")
for val in value:
self.putcell(rownr, val)
rownr += sei[2]
return True
def _repr_html_(self):
"""Give a nice representation of columns in notebooks."""
out="
\n"
# Print column name (not if it is auto-generated)
if not(self.name()[:4]=="Col_"):
out+="
"
out+="
"+self.name()+"
"
out+="
"
cropped=False
rowcount=0
colkeywords=self.getkeywords()
for row in self:
out +="\n
"
out += "
" + _format_cell(row, colkeywords) + "
\n"
out += "
\n"
rowcount+=1
out+="\n"
if rowcount>=20:
cropped=True
break
if out[-2:]=="\n\n":
out=out[:-1]
out+="
"
if cropped:
out+="
("+str(self.nrows()-20)+" more rows)
\n"
return out
python-casacore-2.2.1/casacore/tables/tablehelper.py 0000664 0000000 0000000 00000023002 13203113151 0022471 0 ustar 00root root 0000000 0000000 # tablehelper.py: Helper table functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: tableutil.py,v 1.6 2006/11/08 00:12:55 gvandiep Exp $
from casacore.six import PY2
import numpy
import re
from ..quanta import quantity
# A keywordset in a table can hold tables, but it is not possible to
# pass them around because a ValueHolder cannot deal with it.
# Therefore it is passed around as a string with a special prefix.
def _add_prefix(name):
"""Add the prefix 'Table: ' to a table name to get a specific keyword value."""
return 'Table: ' + name
def _do_remove_prefix(name):
"""Strip the possible prefix 'Table: ' from a table name."""
res = name
if isinstance(res, str):
if (res.find('Table: ') == 0):
res = res.replace('Table: ', '', 1)
return res
def _remove_prefix(name):
"""Strip the possible prefix 'Table: ' from one or more table names."""
if isinstance(name, str):
return _do_remove_prefix(name)
return [_do_remove_prefix(nm) for nm in name]
def _check_index(key, name):
# The __index__ method converts e.g. np.int16 to a proper integer.
# An exception is thrown if the type does not have __index__ which
# means that the given key cannot be used as an index.
try:
return key.__index__()
except:
raise TypeError(name + " indices must be integer (or None in a slice)")
# Check a key or slice given to index a tablerow or tablecolumn object.
# A TypeError exception is raised if values or not integer or None.
# An IndexError is raised if incorrect values are given.
# It returns a list of length 1 if a single index is given.
# Otherwise it returns [startrow, nrow, step].
def _check_key_slice(key, nrows, name):
if not isinstance(key, slice):
inx = _check_index(key, name)
# A single index (possibly negative, thus from the end).
if inx < 0:
inx += nrows
if inx < 0 or inx >= nrows:
raise IndexError(name + " index out of range")
return [inx]
# Given as start:stop:step where each part is optional and can
# be negative.
incr = 1
if key.step is not None:
incr = _check_index(key.step, name)
if incr == 0:
raise RuntimeError(name + " slice step cannot be zero")
strow = 0
endrow = nrows
if incr < 0:
strow = nrows - 1
endrow = -1
if key.start is not None:
strow = _check_index(key.start, name)
if strow < 0:
strow += nrows
strow = min(max(strow, 0), nrows - 1)
if key.stop is not None:
endrow = _check_index(key.stop, name)
if endrow < 0:
endrow += nrows
endrow = min(max(endrow, -1), nrows)
if incr > 0:
nrow = int((endrow - strow + incr - 1) / incr)
else:
nrow = int((strow - endrow - incr - 1) / -incr)
nrow = max(0, nrow)
return [strow, nrow, incr]
# Convert Python value type to a glish-like type string
# as expected by the table code.
def _value_type_name(value):
if isinstance(value, bool):
return 'boolean'
if isinstance(value, int):
return 'integer'
if PY2 and isinstance(value, long):
return 'integer'
if isinstance(value, float):
return 'double'
if isinstance(value, complex):
return 'dcomplex'
if isinstance(value, str):
return 'string'
if isinstance(value, dict):
return 'record'
return 'unknown'
def _format_date(val, unit):
"""
Format dates.
:param val: Value (just the value, not a quantity)
:param unit: Unit. Should be 'rad' or 's'
:return: A string representation of this date.
>>> _format_date(4914741782.503475, 's')
"14-Aug-2014/14:03:03"
"""
if val==numpy.floor(val) and unit=='d':
# Do not show time part if 0
return quantity(val,unit).formatted('YMD_ONLY')
else:
return quantity(val,unit).formatted('DMY')
def _format_quantum(val, unit):
"""
Format a quantity with reasonable units.
:param val: The value (just the value, not a quantity)
:param unit: Unit (something that can be fed to quanta).
:return: A string representation of this quantity.
>>> _format_quantum(3, 'm')
"3 m"
>>> _format_quantum(4914741782.503475, 's')
"4.91474e+09 s"
"""
q=quantity(val,unit)
if q.canonical().get_unit() in ['rad','s']:
return quantity(val, 'm').formatted()[:-1]+unit
else:
return q.formatted()
def _format_cell(val, colkeywords):
"""
Format a cell of the table. Colkeywords can add units.
:param val: A plain value (not a quantum)
:param colkeywords:
:return: A HTML representation of this cell.
"""
out=""
# String arrays are returned as dict, undo that for printing
if isinstance(val, dict):
tmpdict=numpy.array(val['array'])
tmpdict.reshape(val['shape'])
# Leave out quotes around strings
numpy.set_printoptions(formatter={'all':lambda x: str(x)})
out+=numpy.array2string(tmpdict, separator=', ')
# Revert previous numpy print options
numpy.set_printoptions(formatter=None)
else:
valtype='other'
# Check if the column unit is like 'm' or ['m','m','m']
singleUnit=('QuantumUnits' in colkeywords and
(numpy.array(colkeywords['QuantumUnits'])==numpy.array(colkeywords['QuantumUnits'])[0]).all())
if colkeywords.get('MEASINFO',{}).get('type')=='epoch' and singleUnit:
# Format a date/time. Use quanta for scalars, use numpy for array logic around it
# (quanta does not support higher dimensional arrays)
valtype='epoch'
if isinstance(val, numpy.ndarray):
numpy.set_printoptions(formatter={'all':lambda x: _format_date(x,colkeywords['QuantumUnits'][0])})
out+=numpy.array2string(val,separator=', ')
numpy.set_printoptions(formatter=None)
else:
out+=_format_date(val,colkeywords['QuantumUnits'][0])
elif colkeywords.get('MEASINFO',{}).get('type')=='direction' and singleUnit and val.shape==(1,2):
# Format one direction. TODO: extend to array of directions
valtype='direction'
out+="["
part=quantity(val[0,0],'rad').formatted("TIME",precision=9)
part=re.sub(r'(\d+):(\d+):(.*)',r'\1h\2m\3',part)
out+=part+", "
part=quantity(val[0,1],'rad').formatted("ANGLE",precision=9)
part=re.sub(r'(\d+)\.(\d+)\.(.*)',r'\1d\2m\3',part)
out+=part+"]"
elif isinstance(val, numpy.ndarray) and singleUnit:
# Format any array with units
valtype='quanta'
numpy.set_printoptions(formatter={'all':lambda x: _format_quantum(x, colkeywords['QuantumUnits'][0])})
out+=numpy.array2string(val,separator=', ')
numpy.set_printoptions(formatter=None)
elif isinstance(val, numpy.ndarray):
valtype='other'
# Undo quotes around strings
numpy.set_printoptions(formatter={'all':lambda x: str(x)})
out+=numpy.array2string(val,separator=', ')
numpy.set_printoptions(formatter=None)
elif singleUnit:
valtype='onequantum'
out+=_format_quantum(val, colkeywords['QuantumUnits'][0])
else:
valtype='other'
out+=str(val)
if 'QuantumUnits' in colkeywords and valtype=='other':
# Print units if they haven't been taken care of
if not (numpy.array(colkeywords['QuantumUnits'])==numpy.array(colkeywords['QuantumUnits'])[0]).all():
# Multiple different units for element in an array.
# For now, just print the units and let the user figure out what it means
out+=" "+str(colkeywords['QuantumUnits'])
else:
out+=" "+colkeywords['QuantumUnits'][0]
# Numpy sometimes adds double newlines, don't do that
out=out.replace('\n\n','\n')
return out
def _format_row(row, colnames, tab):
"""
Helper function for _repr_html. Formats one row.
:param row: row of this table
:param colnames: vector of column names
:param tab: table, used to get the column keywords
:return: html-formatted row
"""
out=""
out+="\n
\n"
return out
python-casacore-2.2.1/casacore/tables/tableindex.py 0000664 0000000 0000000 00000016573 13203113151 0022340 0 ustar 00root root 0000000 0000000 # tableindex.py: Python tableindex functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: tableindex.py,v 1.6 2006/11/08 00:12:55 gvandiep Exp $
# Make interface to class TableIndexProxy available.
from ._tables import TableIndex
class tableindex(TableIndex):
"""The Python interface to Casacore table index.
A tableindex makes it possible to find rows in a :class:`table` based on
the contents of one or more columns. When constructing the `tableindex` it
has to be specified for which column or columns an index has to be built.
Those columns will be loaded in memory and thereafter row numbers can be
found in a fast way using a binary search.
Using a table index is only useful if many searches will be done in the
table. For a single or few searches it is better to query the table using
method :func:`table.query`.
Normally an index will be build on one or more scalar columns (e.g.
on ANTENNA1 and ANTENNA2 in a measurementset table). However, it is also
possible to buo.d an index for a column containing arrays (e.g. for a
column where each cell can contain multiple names. In that case only a
single column can be indexed.
The index can be unique, but does not need to be.
A unique index can be asked for the row number containing a given key.
A non-unique index can only be asked for the row numbers containing a key.
The returned sequence can thereafter be used in :func:`table.selectrows`
to form that subset of the table.
`tableindex` supports Python's index operator [] as explained in the
methods :func:`rownr` and :func:`rownrs`.
"""
def __init__(self, table, columnnames, sort=True):
TableIndex.__init__(self, table, columnnames, not sort)
"""Create the index on one or more columns.
By default the columns get sorted when forming in the index. By giving
`sort=False` this can be omitted in case the table is already in the
correct order.
Method :func:`table.index` is a somewhat easier way to create a
`tableindex` object.
"""
# Turn a key into a dict if needed.
def _makekey(self, key):
d = key
if not isinstance(d, dict):
cols = self.colnames()
if len(cols) != 1:
raise RuntimeError("key has to be given as a dict for a multi-column index")
d = {cols[0]: key}
return d
def rownr(self, key):
"""Get the unique row number containing the key.
If the index is made from a single column, the keycan be given as a
single value.
Otherwise the key has to be given as a dict where the name of each
field in the dict should correspond with the column name in the index.
For example::
t = table('3c343.MS/ANTENNA')
tinx = t.index ('NAME') # build index for antenna name
rownr = tinx.rownr('RTE') # find an antenna by name
rownr = tinx['RTE'] # same as above
t.getcell ('POSITION', rownr) # get position of that antenna
As shown in the example above the python index operator can also
be used to find a row number if the index if made of a single column.
An exception will be raised if the index is not unique. In that case
method :func:`rownrs` should be used instead.
"""
return self._rownr(self._makekey(key))
def rownrs(self, key, upperkey={}, lowerincl=True, upperincl=True):
"""Get a sequence of row numbers containing the key(s).
A single key can be given, but by giving argument `upperkey` as well
a key range can be given (where upper key must be > lower).
One can specify if the lower and upper key should be part of the range
(`incl=True`) or not. By default both keys are part of the range.
The key and optional upper key have to be given in the same way as
for method :func:`rownr`.
Similar to method :func:`rownr`. python's index operator [] can be used
if the index consists of a single column. However, in this case only
key ranges can be used (because the index operator with a single key
returns a single row number, thus can only be used for unique indices).
The lower key is inclusive, but the upper key is exclusive conform
the standard python index semantics.
For example::
t = table('3c343.MS')
tinx = t.index ('ANTENNA1') # build index for antenna name
rownr = tinx.rownr(0) # find antenna1 = 0
rownr = tinx[0:1] # same as above
"""
lkey = self._makekey(key)
ukey = self._makekey(upperkey)
if len(ukey) == 0:
return self._rownrs(lkey)
return self._rownrsrange(lkey, ukey, lowerincl, upperincl)
def isunique(self):
"""Tell if all keys in the index are unique."""
return self._isunique()
def colnames(self):
"""Return the column names the index is made of."""
return self._colnames()
def setchanged(self, columnnames=[]):
"""Tell the index that data has changed.
The index is smart enough to detect that the number of rows in the
indexed table has changed. However, it cannot detect if a value in
a column contained in this inex has changed. So it has to be told
explicitly.
`columnnames`
The names of the columns in which data have changed.
Giving no names means that all columns in the index have changed.
"""
return self._setchanged(columnnames)
def __getitem__(self, key):
if not isinstance(key, slice):
rnr = self.rownr(key)
if rnr < 0:
raise KeyError("key not found in tableindex")
return rnr
if key.step is not None:
raise RuntimeError("tableindex slicing cannot have a step")
lowerkey = 0
if key.start is not None:
lowerkey = key.start
upperkey = 2147483647; # highest int
if key.stop is not None:
upperkey = key.stop
if (lowerkey >= upperkey):
raise RuntimeError("tableindex slice stop must be > start")
rnrs = self.rownrs(lowerkey, upperkey, True, False)
if len(rnrs) == 0:
raise KeyError("keys not found in tableindex")
return rnrs
python-casacore-2.2.1/casacore/tables/tableiter.py 0000664 0000000 0000000 00000005105 13203113151 0022161 0 ustar 00root root 0000000 0000000 # tableiter.py: Python tableiter functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: tableiter.py,v 1.6 2006/12/11 02:46:08 gvandiep Exp $
# Make interface to class TableIterProxy available.
from ._tables import TableIter
from .table import table
class tableiter(TableIter):
"""The Python interface to Casacore table iterators
A `tableiter` allows iteration through a table based on the contents
of one or more columns. Each step in the iteration process forms
a subset of the table for which the specified columns have the same value.
It can easily be constructed using the :func:`table.iter` method as done
in the example below::
t = table('3c343.MS')
for ts in t.iter('ANTENNA1'):
print ts.nrows()
In this example `ts` will be a so-called reference table which can be
operated on like any other table object.
Multiple column names should be given in a sequence (tuple or list).
"""
def __init__(self, table, columnnames, order='', sort=True):
st = sort
if isinstance(sort, bool):
st = 'heapsort'
if not sort:
st = 'nosort'
TableIter.__init__(self, table, columnnames, order, st)
def __iter__(self):
# __iter__ is needed
return self
def next(self):
# next returns a Table object, so turn that into table.
return table(self._next(), _oper=3)
def reset(self):
"""Reset the iterator to the beginning."""
self._reset()
__next__ = next
python-casacore-2.2.1/casacore/tables/tablerow.py 0000664 0000000 0000000 00000013101 13203113151 0022020 0 ustar 00root root 0000000 0000000 # tablerow.py: Python tablerow functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id: tablerow.py,v 1.6 2007/08/28 07:22:18 gvandiep Exp $
# Make interface to class TableRowProxy available.
from ._tables import TableRow
from .tablehelper import _check_key_slice
# A normal tablerow object keeps a reference to a table object to be able
# to know the actual number of rows.
# However, a mutual dependency is created when doing that for the tablerow
# object inside the table object.
# Therefore an intermediate _tablerow exists to be used in class table.
class _tablerow(TableRow):
def __init__(self, table, columnnames, exclude=False):
TableRow.__init__(self, table, columnnames, exclude)
def iswritable(self):
"""Tell if all columns in the row object are writable."""
return self._iswritable()
def get(self, rownr):
"""Get the contents of the given row."""
return self._get(rownr)
def put(self, rownr, value, matchingfields=True):
"""Put the values into the given row.
The value should be a dict (as returned by method :func:`get`.
The names of the fields in the dict should match the names of the
columns used in the `tablerow` object.
`matchingfields=True` means that the value may contain more fields
and only fields matching a column name will be used.
"""
self._put(rownr, value, matchingfields)
def _getitem(self, key, nrows):
sei = _check_key_slice(key, nrows, 'tablerow')
rownr = sei[0]
if len(sei) == 1:
return self.get(rownr)
result = []
inx = 0
while inx < sei[1]:
result.append(self.get(rownr))
rownr += sei[2]
inx += 1
return result
def _setitem(self, key, value, nrows):
sei = _check_key_slice(key, nrows, 'tablerow')
rownr = sei[0]
if len(sei) == 1:
return self.put(rownr, value)
if isinstance(value, dict):
# The same value is put in all rows.
inx = 0
while inx < sei[1]:
self.put(rownr, value, True)
rownr += sei[2]
inx += 1
else:
# Each row has its own value.
if len(value) != sei[1]:
raise RuntimeError("tablerow slice length differs from value length")
for val in value:
self.put(rownr, val, True)
rownr += sei[2]
class tablerow(_tablerow):
"""The Python interface to Casacore table rows.
A table row is a record (dict) containing the values of a single row for
one or more columns in a table. In constructing the `tablerow` object, one
can specify which columns are to be included or excluded.
By default all columns will be used, but if the table is writable,
only writable columns will be used.
A `tablerow` object can easily be constructed using :func:`table.row`.
One or more rows can be read or written using the standard python indexing
syntax where (negative) strides are possible.
For example:
t = table ('3c343.MS')
tr = t.row (['ANTENNA1', 'ANTENNA2', 'ARRAY_ID'])
tr[0] # get row 0
tr[:5] # get row 0,1,2,3,4
tr[-5,-1,] # get last 4 rows
tr[-1,-5,-1] # get last 4 rows in reversed order
tr[1] = tr[0] # put values of row 0 into row 1
Note that the last line will fail because the table is opened readonly.
The argument `readonly=False` is needed in the table constructor to make
it work.
The `tablerow` class supports the context manager idiom (__enter__ and __exit__).
When used in a `with` statement, the table changes will be flushed
automatically, which is handy when writing to table rows.
For example::
with t.row() as tr:
tr.put (1, tr.get(0)) # copy row 0 to row 1
"""
def __init__(self, table, columnnames=[], exclude=False):
_tablerow.__init__(self, table, columnnames, exclude)
self._table = table
def __enter__(self):
"""Function to enter a with block."""
return self
def __exit__(self, type, value, traceback):
"""Function to exit a with block which flushes the table object."""
self._table.flush()
def __len__(self):
return self._table.nrows()
def __getitem__(self, key):
return self._getitem(key, self._table.nrows())
def __setitem__(self, key, value):
return self._setitem(key, value, self._table.nrows())
python-casacore-2.2.1/casacore/tables/tableutil.py 0000775 0000000 0000000 00000066674 13203113151 0022220 0 ustar 00root root 0000000 0000000 # tableutil.py: Utility table functions
# Copyright (C) 2006
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
from collections import defaultdict
from casacore import six
from .table import table
from .tablehelper import _remove_prefix, _value_type_name
def tablefromascii(tablename, asciifile,
headerfile='',
autoheader=False, autoshape=[],
columnnames=[], datatypes=[],
sep=' ',
commentmarker='',
firstline=1, lastline=-1,
readonly=True,
lockoptions='default', ack=True):
"""Create a table from an ASCII file.
Create a table from a file in ASCII format. Columnar data as well as
table and column keywords may be specified.
Once the table is created from the ASCII data, it is opened in the
specified mode and a table object is returned.
The table columns are filled from a file containing the data values
separated by a separator (one line per table row). The default
separator is a blank. Blanks before and after the separator are ignored.
If a non-blank separator is used, values can be empty. Such values
default to 0, empty string, or F depending on the data type. E.g.
1,,2, has 4 values of which the 2nd and 4th are empty and default to 0.
Similarly if fewer values are given than needed, the missing values
get the default value.
Either the data format can be explicitly specified or it can be found
automatically. The former gives more control in ambiguous situations.
Both scalar and array columns can be generated from the ASCII input.
The format string determines the type and optional shape.
It is possible to give the column names and their data types in
various ways:
- Using 2 header lines (as described below) as the first two lines
in the data file or in a separate header file. This is the default way.
- Derive them automatically from the data (`autoheader=True`).
- Using the arguments `columnnames` and
`datatypes` (as non-empty vectors of strings).
It implies (`autoheader=False`). The data types should be
given in the same way as done in headers.
In automatic mode (`autoheader=True`) the first line
of the ASCII data is analyzed
to deduce the data types. Only the types I, D, and A can be
recognized. A number without decimal point or exponent is I (integer),
otherwise it is D (double). Any other string is A (string).
Note that a number may contain a leading sign (+ or -).
The `autoshape` argument can be used to specify if the input
should be stored as multiple scalars (the default) or as a single
array. In the latter case one axis in the shape can be defined as
variable length by giving it the value 0. It means that the actual
array shape in a row is determined by the number of values in the
corresponding input line.
Columns get the names `Column1`, `Column2`, etc..
For example:
1. `autoshape=[]` (which is the default) means that all values
are to be stored as scalar columns.
2. `autoshape=0` means that all values in a row are to be stored as
a variable length vector.
3. `autoshape=10` defines a fixed length vector. If an input
line contains less than 10 values, the vector is filled with default
values. If more than 10 values, the latter values are ignored.
4. `autoshape=[5,0]` defines a 2-dim array of which the 2nd axis is
variable. Note that if an input line does not contain a multiple of 5
values, the array is filled with default values.
If the format of the table is explicitly specified, it has to be done
either in the first two lines of the data file (named by the
argument filename), or in a separate header file (named by the
argument headerfile). In both forms, table keywords may also be
specified before the column definitions.
The column names and types can be described by two lines:
1. The first line contains the names of the columns.
These names may be enclosed in quotes (either single or double).
2. The second line contains the data type and optionally the shape
of each column. Valid types are:
- S for Short data
- I for Integer data
- R for Real data
- D for Double Precision data
- X for Complex data (Real followed by Imaginary)
- Z for Complex data (Amplitude then Phase)
- DX for Double Precision Complex data (Real followed by Imaginary)
- DZ for Double Precision Complex data (Amplitude then Phase)
- A for ASCII data (a value must be enclosed in single or double quotes
if it contains whitespace)
- B for Boolean data (False are empty string, 0, or any string
starting with F, f, N, or n).
If a column is an array, the shape has to be given after the data type
without any whitespace. E.g. `I10` defines an integer vector
of length 10. `A2,5` defines a 2-dim string array with shape
[2,5]. Note that `I` is not the same as `I1` as the
first one defines a scalar and the other one a vector with length 1.
The last column can have one variable length axis denoted by the value 0.
It "consumes" the remainder of the input line.
If the argument headerfile is set then the header information is
read from that file instead of the first lines of the data file.
To give a simple example of the form where the header information
is located at the top of the data file::
COLI COLF COLD COLX COLZ COLS
I R D X Z A
1 1.1 1.11 1.12 1.13 1.14 1.15 Str1
10 11 12 13 14 15 16 ""
Note that a complex number consists of 2 numbers.
Also note that an empty string can be given.
Let us now give an example of a separate header file that one might use to
get interferometer data into casacore::
U V W TIME ANT1 ANT2 DATA
R R R D I I X1,0
The data file would then look like::
124.011 54560.0 3477.1 43456789.0990 1 2 4.327 -0.1132
34561.0 45629.3 3900.5 43456789.0990 1 3 5.398 0.4521
Note that the DATA column is defined as a 2-dim array of 1
correlation and a variable number of channels, so the actual number of
channels is determined by the input. In this example both rows will
have 1 channel (note that a complex value contains 2 values).
Tables may have keywords in addition to the columns. The keywords
are useful for holding information that is global to the entire
table (such as author, revision, history, etc.).
The keywords in the header definitions must preceed the column descriptions.
They must be enclosed between a line that starts with ".key..." and
a line that starts with ".endkey..." (where ... can be anything).
A table keywordset and column keywordsets can be specified.
The latter can be specified by specifying the column name after
the .keywords string.
Between these two lines each line should contain the following:
- The keyword name, e.g., ANYKEY
- The datatype and optional shape of the keyword
(cf. list of valid types above)
- The value or values for the keyword (the keyword may contain
a scalar or an array of values). e.g., 3.14159 21.78945
Thus to continue the example above, one might wish to add keywords
as follows::
.keywords
DATE A "97/1/16"
REVISION D 2.01
AUTHOR A "Tim Cornwell"
INSTRUMENT A "VLA"
.endkeywords
.keywords TIME
UNIT A "s"
.endkeywords
U V W TIME ANT1 ANT2 DATA
R R R D I I X1,0
Similarly to the column format string, the keyword formats can also
contain shape information. The only difference is that if no shape is
given, a keyword can have multiple values (making it a vector).
It is possible to ignore comment lines in the header and data file
by giving the `commentmarker`. It indicates that lines
starting with the given marker are ignored. Note that the marker can
be a regular expression (e.g. `' *//'` tells that lines starting
with // and optionally preceeded by blanks have to be ignored).
With the arguments `firstline` and `lastline` one can
specify which lines have to be taken from the input file. A negative value
means 1 for `firstline` or end-of-file for `lastline`.
Note that if the headers and data are combined in one file,
these line arguments apply to the whole file. If headers and data are in
separate files, these line arguments apply to the data file only.
Also note that ignored comment lines are counted, thus are used to
determine which lines are in the line range.
The number of rows is determined by the number of lines read from the data
file.
"""
import os.path
filename = os.path.expandvars(asciifile)
filename = os.path.expanduser(filename)
if not os.path.exists(filename):
s = "File '%s' not found" % (filename)
raise IOError(s)
if headerfile != '':
filename = os.path.expandvars(headerfile)
filename = os.path.expanduser(filename)
if not os.path.exists(filename):
s = "File '%s' not found" % (filename)
raise IOError(s)
tab = table(asciifile, headerfile, tablename, autoheader, autoshape,
sep, commentmarker, firstline, lastline,
_columnnames=columnnames, _datatypes=datatypes, _oper=1)
six.print_('Input format: [' + tab._getasciiformat() + ']')
# Close table and reopen it in correct way.
tab = 0
return table(tablename, readonly=readonly, lockoptions=lockoptions,
ack=ack)
# Create a description of a scalar column
def makescacoldesc(columnname, value,
datamanagertype='',
datamanagergroup='',
options=0, maxlen=0, comment='',
valuetype='', keywords={}):
"""Create description of a scalar column.
A description for a scalar column can be created from a name for
the column and a data value, which is used only to determine the
type of the column. Note that a dict value is also possible.
It is possible to create the column description in more detail
by giving the data manager name, group, option, and comment as well.
The data manager type tells which data manager (storage manager)
is used to store the columns. The data manager type and group are
explained in more detail in the `casacore Tables
<../../casacore/doc/html/group__Tables__module.html>`_ documentation.
It returns a dict with fields `name` and `desc` which can thereafter be used
to build a table description using function :func:`maketabdesc`.
`columname`
Name of column
`value`
Example data value used to determine the column's data type.
It is only used if argument `valuetype` is not given.
`datamanagertype`
Type of data manager which can be one of StandardStMan (default)
or IncrementalStMan. The latter one can save disk space if many subsequent
cells in the column will have the same value.
`datamanagergroup`
Data manager group. Only for the expert user.
`options`
Options. Need not be filled in.
`maxlen`
Maximum length of string values in a column.
Default 0 means unlimited.
`comment`
Comment: informational for user.
`valuetype`
A string giving the column's data type. Possible data types are
bool (or boolean), uchar (or byte), short, int (or integer), uint,
float, double, complex, dcomplex, and string.
'keywords'
A dict defining initial keywords for the column.
For example::
scd1 = makescacoldesc("col2", ""))
scd2 = makescacoldesc("col1", 1, "IncrementalStMan")
td = maketabdesc([scd1, scd2])
This creates a table description consisting of an integer column `col1`,
and a string column `col2`. `col1` uses the IncrementalStMan storage manager,
while `col2` uses the default storage manager StandardStMan.
"""
vtype = valuetype
if vtype == '':
vtype = _value_type_name(value)
rec2 = {'valueType': vtype,
'dataManagerType': datamanagertype,
'dataManagerGroup': datamanagergroup,
'option': options,
'maxlen': maxlen,
'comment': comment,
'keywords': keywords}
return {'name': columnname,
'desc': rec2}
# Create a description of an array column
def makearrcoldesc(columnname, value, ndim=0,
shape=[], datamanagertype='',
datamanagergroup='',
options=0, maxlen=0, comment='',
valuetype='', keywords={}):
"""Create description of an array column.
A description for a scalar column can be created from a name for
the column and a data value, which is used only to determine the
type of the column. Note that a dict value is also possible.
It is possible to create the column description in more detail
by giving the dimensionality, shape, data manager name, group, option,
and comment as well.
The data manager type tells which data manager (storage manager)
is used to store the columns. The data manager type and group are
explained in more detail in the `casacore Tables
<../../casacore/doc/html/group__Tables__module.html>`_ documentation.
It returns a dict with fields `name` and `desc` which can thereafter be used
to build a table description using function :func:`maketabdesc`.
`name`
The name of the column.
`value`
A data value, which is only used to determine the data type of the column.
It is only used if argument `valuetype` is not given.
`ndim`
Optionally the number of dimensions. A value > 0 means that all
arrays in the column must have that dimensionality. Note that the
arrays can still differ in shape unless the shape vector is also given.
`shape`
An optional sequence of integers giving the shape of the array in each
cell. If given, it forces option FixedShape (see below) and sets the
number of dimensions (if not given). All arrays in the column get the
given shape and the array is created as soon as a row is added.
Note that the shape vector gives the shape in each table cell; the
number of rows in the table should NOT be part of it.
`datamanagertype`
Type of data manager which can be one of StandardStMan (default),
IncrementalStMan, TiledColumnStMan, TiledCellStMan, or TiledShapeStMan.
The tiled storage managers are usually used for bigger data arrays.
`datamanagergroup`
Data manager group. Only for the expert user.
`options`
Optionally numeric array options which can be added to combine them.
`1` means Direct.
It tells that the data are directly stored in the table. Direct
forces option FixedShape. If not given, the array is indirect, which
means that the data will be stored in a separate file.
`4` means FixedShape.
This option does not need to be given, because it is enforced if
the shape is given. FixedShape means that the shape of the array must
be the same in each cell of the column. Otherwise the array shapes may
be different in each column cell and is it possible that a cell does
not contain an array at all.
Note that when given (or implicitly by option Direct), the
shape argument must be given as well.
Default is 0, thus indirect and variable shaped.
`maxlen`
Maximum length of string values in a column.
Default 0 means unlimited.
`comment`
Comment: informational for user.
`valuetype`
A string giving the column's data type. Possible data types are
bool (or boolean), uchar (or byte), short, int (or integer), uint,
float, double, complex, dcomplex, and string.
'keywords'
A dict defining initial keywords for the column.
For example::
acd1= makescacoldesc("arr1", 1., 0, [2,3,4])
td = maketabdesc(acd1)
This creates a table description consisting of an array column `arr1`
containing 3-dim arrays of doubles with shape [2,3,4].
"""
vtype = valuetype
if vtype == '':
vtype = _value_type_name(value)
if len(shape) > 0:
if ndim <= 0:
ndim = len(shape)
rec2 = {'valueType': vtype,
'dataManagerType': datamanagertype,
'dataManagerGroup': datamanagergroup,
'ndim': ndim,
'shape': shape,
'_c_order': True,
'option': options,
'maxlen': maxlen,
'comment': comment,
'keywords': keywords}
return {'name': columnname,
'desc': rec2}
# Create a description of a column
def makecoldesc(columnname, desc):
"""Create column description using the description of another column.
The other description can be obtained from a table using function
:func:`getcoldesc` or from another column description dict using
`otherdesc['desc']`.
It returns a dict with fields `name` and `desc` which can thereafter be used
to build a table description using function :func:`maketabdesc`.
`columname`
Name of column
`desc`
Description of the column
For example::
cd1 = makecoldesc("col2", t.getcoldesc('othercol'))
td = maketabdesc(cd1)
This creates a table description consisting of a column `col2` having
the same description as column `othercol`.
"""
return {'name': columnname,
'desc': desc}
# Create a table description from a set of column descriptions
def maketabdesc(descs=[]):
"""Create a table description.
Creates a table description from a set of column descriptions. The
resulting table description can be used in the :class:`table` constructor.
For example::
scd1 = makescacoldesc("col2", "aa")
scd2 = makescacoldesc("col1", 1, "IncrementalStMan")
scd3 = makescacoldesc("colrec1", {})
acd1 = makearrcoldesc("arr1", 1, 0, [2,3,4])
acd2 = makearrcoldesc("arr2", 0.+0j)
td = maketabdesc([scd1, scd2, scd3, acd1, acd2])
t = table("mytable", td, nrow=100)
| This creates a table description `td` from five column descriptions
and then creates a 100-row table called `mytable` from the table
description.
| The columns contain respectivily strings, integer scalars, records,
3D integer arrays with fixed shape [2,3,4], and complex arrays with
variable shape.
"""
rec = {}
# If a single dict is given, make a list of it.
if isinstance(descs, dict):
descs = [descs]
for desc in descs:
colname = desc['name']
if colname in rec:
raise ValueError('Column name ' + colname + ' multiply used in table description')
rec[colname] = desc['desc']
return rec
def makedminfo(tabdesc, group_spec=None):
"""Creates a data manager information object.
Create a data manager information dictionary outline from a table description.
The resulting dictionary is a bare outline and is available for the purposes of
further customising the data manager via the `group_spec` argument.
The resulting dictionary can be used in the :class:`table` constructor and
the :meth:`default_ms` and :meth:`default_ms_subtable` functions.
`tabdesc`
The table description
`group_spec`
The SPEC for a data manager group. In practice this is useful for
setting the Default Tile Size and Maximum Cache Size for the Data Manager
{
'WeightColumnGroup' : {
'DEFAULTTILESHAPE': np.int32([4,4,4]),
'MAXIMUMCACHESIZE': 1000,
}
}
This should be used with care.
"""
if group_spec is None:
group_spec = {}
class DMGroup(object):
"""
Keep track of the columns, type and spec of each data manager group
"""
def __init__(self):
self.columns = []
self.type = None
self.spec = None
dm_groups = defaultdict(DMGroup)
# Iterate through the table columns, grouping them
# by their dataManagerGroup
for c, d in six.iteritems(tabdesc):
if c in ('_define_hypercolumn_', '_keywords_', '_private_keywords_'):
continue
# Extract group and data manager type
group = d.get("dataManagerGroup", "StandardStMan")
type_ = d.get("dataManagerType", "StandardStMan")
# Set defaults if necessary
if not group:
group = "StandardStMan"
if not type_:
type_ = "StandardStMan"
# Obtain the (possibly empty) data manager group
dm_group = dm_groups[group]
# Add the column
dm_group.columns.append(c)
# Set the spec
if dm_group.spec is None:
dm_group.spec = group_spec.get(group, {})
# Check that the data manager type is consistent across columns
if dm_group.type is None:
dm_group.type = type_
elif not dm_group.type == type_:
raise ValueError("Mismatched dataManagerType '%s' "
"for dataManagerGroup '%s' "
"Previously, the type was '%s'" %
(type_, group, dm_group.type))
# Output a data manager entry
return {
'*%d'%(i+1): {
'COLUMNS': dm_group.columns,
'TYPE': dm_group.type,
'NAME': group,
'SPEC' : dm_group.spec,
'SEQNR': i
} for i, (group, dm_group)
in enumerate(six.iteritems(dm_groups))
}
# Create the old glish names for them.
tablecreatescalarcoldesc = makescacoldesc
tablecreatearraycoldesc = makearrcoldesc
tablecreatedesc = maketabdesc
tablecreatedm = makedminfo
# Define a hypercolumn in the table description.
def tabledefinehypercolumn(tabdesc,
name, ndim, datacolumns,
coordcolumns=False,
idcolumns=False):
"""Add a hypercolumn to a table description.
It defines a hypercolumn and adds it the given table description.
A hypercolumn is an entity used by the Tiled Storage Managers (TSM). It
defines which columns have to be stored together with a TSM.
It should only be used by expert users who want to use a TSM to its
full extent. For a basic TSM s hypercolumn definition is not needed.
tabledesc
A table description (result from :func:`maketabdesc`).
name
Name of hypercolumn
ndim
Dimensionality of hypercolumn; normally 1 more than the dimensionality
of the arrays in the data columns to be stored with the TSM
datacolumns
Data columns to be stored with TSM
coordcolumns
Optional coordinate columns to be stored with TSM
idcolumns
Optional id columns to be stored with TSM
For example::
scd1 = makescacoldesc("col2", "aa")
scd2 = makescacoldesc("col1", 1, "IncrementalStMan")
scd3 = makescacoldesc("colrec1", {})
acd1 = makearrcoldesc("arr1", 1, 0, [2,3,4])
acd2 = makearrcoldesc("arr2", as_complex(0))
td = maketabdesc([scd1, scd2, scd3, acd1, acd2])
tabledefinehypercolumn(td, "TiledArray", 4, ["arr1"])
tab = table("mytable", tabledesc=td, nrow=100)
| This creates a table description `td` from five column descriptions
and then creates a 100-row table called mytable from the table
description.
| The columns contain respectivily strings, integer scalars, records,
3D integer arrays with fixed shape [2,3,4], and complex arrays with
variable shape.
| The first array is stored with the Tiled Storage Manager (in this case
the TiledColumnStMan).
"""
rec = {'HCndim': ndim,
'HCdatanames': datacolumns}
if not isinstance(coordcolumns, bool):
rec['HCcoordnames'] = coordcolumns
if not isinstance(idcolumns, bool):
rec['HCidnames'] = idcolumns
if '_define_hypercolumn_' not in tabdesc:
tabdesc['_define_hypercolumn_'] = {}
tabdesc['_define_hypercolumn_'][name] = rec
def tabledelete(tablename, checksubtables=False, ack=True):
"""Delete a table on disk.
It is the same as :func:`table.delete`, but without the need to open
the table first.
"""
tabname = _remove_prefix(tablename)
t = table(tabname, ack=False)
if t.ismultiused(checksubtables):
six.print_('Table', tabname, 'cannot be deleted; it is still in use')
else:
t = 0
table(tabname, readonly=False, _delete=True, ack=False)
if ack:
six.print_('Table', tabname, 'has been deleted')
def tableexists(tablename):
"""Test if a table exists."""
result = True
try:
t = table(tablename, ack=False)
except:
result = False
return result
def tableiswritable(tablename):
"""Test if a table is writable."""
result = True
try:
t = table(tablename, readonly=False, ack=False)
result = t.iswritable()
except:
result = False
return result
def tablecopy(tablename, newtablename, deep=False, valuecopy=False, dminfo={},
endian='aipsrc', memorytable=False, copynorows=False):
"""Copy a table.
It is the same as :func:`table.copy`, but without the need to open
the table first.
"""
t = table(tablename, ack=False)
return t.copy(newtablename, deep=deep, valuecopy=valuecopy,
dminfo=dminfo, endian=endian, memorytable=memorytable,
copynorows=copynorows)
def tablerename(tablename, newtablename):
"""Rename a table.
The table with the given name is renamed (or moved) to the new name.
"""
t = table(tablename, ack=False)
t.rename(newtablename)
def tableinfo(tablename):
"""Get type info of a table.
It is the same as :func:`table.info`, but without the need to open
the table first.
"""
t = table(tablename, ack=False)
return t.info()
def tablesummary(tablename):
"""Get the summary of a table.
It is the same as :func:`table.summary`, but without the need to open
the table first.
"""
t = table(tablename, ack=False)
t.summary()
def tablestructure(tablename, dataman=True, column=True, subtable=False,
sort=False):
"""Print the structure of a table.
It is the same as :func:`table.showstructure`, but without the need to open
the table first.
"""
t = table(tablename, ack=False)
six.print_(t.showstructure(dataman, column, subtable, sort))
python-casacore-2.2.1/casacore/tables/wxtablebrowser.py 0000775 0000000 0000000 00000006617 13203113151 0023274 0 ustar 00root root 0000000 0000000 from wxPython.grid import *
from wxPython.wx import *
from casacore import six
class wxCasaTable(wxPyGridTableBase):
"""
This is all it takes to make a custom data table to plug into a
wxGrid. There are many more methods that can be overridden, but
the ones shown below are the required ones. This table simply
provides strings containing the row and column values.
"""
def __init__(self, log, ctable):
wxPyGridTableBase.__init__(self)
self.log = log
self.casatab = ctable
self.odd = wxGridCellAttr()
self.odd.SetBackgroundColour("gray90")
self.even = wxGridCellAttr()
self.even.SetBackgroundColour("white")
def GetAttr(self, row, col, kind):
attr = [self.even, self.odd][row % 2]
attr.IncRef()
return attr
def GetColLabelValue(self, col):
colnames = self.casatab.colnames()
return colnames[col]
def GetNumberRows(self):
return self.casatab.nrows()
def GetNumberCols(self):
return self.casatab.ncols()
def IsEmptyCell(self, row, col):
return False
def GetValue(self, row, col):
coln = self.casatab.colnames()
cell = 'array'
## if self.casatab.isscalarcol(coln[col]):
## cellval = self.casatab.getcell(coln[col],row)
## if isinstance(cellval,float):
## if coln[col] == "TIME":
## cell = str(cellval)
## else:
## cell = "%3.5f" % cellval
## else:
## cell = str(cellval)
## else:
## cell += self.casatab.getcolshapestring(coln[col],row,1)[0]
## return cell
return str(self.casatab.getcell(coln[col], row))
def SetValue(self, row, col, value):
self.log.write('SetValue(%d, %d, "%s") ignored.\n' % (row, col, value))
# ---------------------------------------------------------------------------
class wxCasaTableGrid(wxGrid):
def __init__(self, parent, log, ctable):
wxGrid.__init__(self, parent, -1)
table = wxCasaTable(log, ctable)
# The second parameter means that the grid is to take ownership of the
# table and will destroy it when done. Otherwise you would need to keep
# a reference to it and call it's Destroy method later.
self.SetTable(table, True)
EVT_GRID_CELL_RIGHT_CLICK(self, self.OnRightDown) # added
def OnRightDown(self, event): # added
six.print_(self.GetSelectedRows()) # added
# ---------------------------------------------------------------------------
class CasaTestFrame(wxFrame):
def __init__(self, parent, log, ctable):
wxFrame.__init__(self, parent, -1, "Casa Table Browser",
size=(640, 480))
grid = wxCasaTableGrid(self, log, ctable)
grid.EnableEditing(False)
# grid.AutoSizeColumns()
# ---------------------------------------------------------------------------
if __name__ == '__main__':
import sys
app = wxPySimpleApp()
from casacore.tables import table as casatable
casatab = casatable('/nfs/aips++/data/atnf/scripts/C972.ms')
frame = CasaTestFrame(None, sys.stdout, casatab)
frame.Show(True)
app.MainLoop()
# ---------------------------------------------------------------------------
python-casacore-2.2.1/casacore/util/ 0000775 0000000 0000000 00000000000 13203113151 0017336 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/casacore/util/__init__.py 0000664 0000000 0000000 00000000143 13203113151 0021445 0 ustar 00root root 0000000 0000000 """
Utilities for casacore modules.
"""
from .substitute import substitute, getlocals, getvariable
python-casacore-2.2.1/casacore/util/substitute.py 0000664 0000000 0000000 00000023343 13203113151 0022130 0 ustar 00root root 0000000 0000000 # substitute.py: substitute python variables and expressions
# Copyright (C) 1998,1999,2008
# Associated Universities, Inc. Washington DC, USA.
#
# This library is free software; you can redistribute it and/or modify it
# under the terms of the GNU Library General Public License as published by
# the Free Software Foundation; either version 2 of the License, or (at your
# option) any later version.
#
# This library is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public
# License for more details.
#
# You should have received a copy of the GNU Library General Public License
# along with this library; if not, write to the Free Software Foundation,
# Inc., 675 Massachusetts Ave, Cambridge, MA 02139, USA.
#
# Correspondence concerning AIPS++ should be addressed as follows:
# Internet email: aips2-request@nrao.edu.
# Postal address: AIPS++ Project Office
# National Radio Astronomy Observatory
# 520 Edgemont Road
# Charlottesville, VA 22903-2475 USA
#
# $Id$
import numpy as np
__all__ = ['getlocals', 'getvariable', 'substitute']
def getlocals(back=2):
"""Get the local variables some levels back (-1 is top)."""
import inspect
fr = inspect.currentframe()
try:
while fr and back != 0:
fr1 = fr
fr = fr.f_back
back -= 1
except:
pass
return fr1.f_locals
def getvariable(name):
"""Get the value of a local variable somewhere in the call stack."""
import inspect
fr = inspect.currentframe()
try:
while fr:
fr = fr.f_back
vars = fr.f_locals
if name in vars:
return vars[name]
except:
pass
return None
def substitute(s, objlist=(), globals={}, locals={}):
"""Substitute global python variables in a command string.
This function parses a string and tries to substitute parts like
`$name` by their value. It is uses by :mod:`image` and :mod:`table`
to handle image and table objects in a command, but also other
variables (integers, strings, etc.) can be substituted.
The following rules apply:
1. A name must start with an underscore or alphabetic, followed
by zero or more alphanumerics and underscores.
2. String parts enclosed in single or double quotes are literals and
are left untouched.
Furthermore a $ can be escaped by a backslash, which is useful
if an environment variable is used. Note that an extra backslash
is required in Python to escape the backslash.
The output contains the quotes and backslashes.
3. A variable is looked up in the given local and global namespaces.
4. If the variable `name` has a vector value, its substitution is
enclosed in square brackets and separated by commas.
5. A string value is enclosed in double quotes. If the value
contains a double quote, that quote is enclosed in single quotes.
6. If the name's value has a type mentioned in the argument `objlist`,
it is substituted by `$n` (where n is a sequence number) and its
value is added to the objects of that type in `objlist`.
7. If the name is unknown or has an unknown type, it is left untouched.
The `objlist` argument is a list of tuples or lists where each tuple
or list has three fields:
1. The first field is the object type (e.g. `table`)
2. The second field is a prefix for the sequence number (usually empty).
E.g. regions could have prefix 'r' resulting in a substitution like
`$r1`.
3. The third field is a list of objects to be substituted. New objects
get appended to it. Usually the list is initially empty.
Apart from substituting variables, it also substitutes `$(expression)`
by the expression result.
It correctly handles parentheses and quotes in the expression.
For example::
>>> a = 2
>>> b = 3
>>> substitute('$(a+b)+$a')
'5+2'
>>> substitute('$(a+b+a)')
'7'
>>> substitute('$((a+b)+$a)')
'$((a+b)+$a)'
>>> substitute('$((a+b)*(a+b))')
'25'
>>> substitute('$(len("ab cd( de"))')
'9'
Substitution is NOT recursive. E.g. if a=1 and b="$a",
the result of substitute("$b") is "$a" and not 1.
"""
# Get the local variables at the caller level if not given.
if not locals:
locals = getlocals(3)
# Initialize some variables.
backslash = False
dollar = False
nparen = 0
name = ''
evalstr = ''
squote = False
dquote = False
out = ''
# Loop through the entire string.
for tmp in s:
if backslash:
out += tmp
backslash = False
continue
# If a dollar is found, we might have a name or expression.
# Alphabetics and underscore are always part of name.
if dollar and nparen == 0:
if tmp == '_' or ('a' <= tmp <= 'z') or ('A' <= tmp <= 'Z'):
name += tmp
continue
# Numerics are only part if not first character.
if '0' <= tmp <= '9' and name != '':
name += tmp
continue
# $( indicates the start of an expression to evaluate.
if tmp == '(' and name == '':
nparen = 1
evalstr = ''
continue
# End of name found. Try to substitute.
out += substitutename(name, objlist, globals, locals)
dollar = False
# Handle possible single or double quotes.
if tmp == '"' and not squote:
dquote = not dquote
elif tmp == "'" and not dquote:
squote = not squote
if not dquote and not squote:
# Count the number of balanced parentheses
# (outside quoted strings) in the subexpression.
if nparen > 0:
if tmp == '(':
nparen += 1
elif tmp == ')':
nparen -= 1
if nparen == 0:
# The last closing parenthese is found.
# Evaluate the subexpression.
# Add the result to the output.
out += substituteexpr(evalstr, globals, locals)
dollar = False
evalstr += tmp
continue
# Set a switch if we have a dollar (outside quoted
# and eval strings).
if tmp == '$':
dollar = True
name = ''
continue
# No special character; add it to output or evalstr.
# Set a switch if we have a backslash.
if nparen == 0:
out += tmp
else:
evalstr += tmp
if tmp == '\\':
backslash = True
# The entire string has been handled.
# Substitute a possible last name.
# Insert a possible incomplete eval string as such.
if dollar:
out += substitutename(name, objlist, globals, locals)
else:
if nparen > 0:
out += '$(' + evalstr
return out
# This function tries to substitute the given name using
# the rules described in the description of function substitute.
def substitutename(name, objlist, globals, locals):
# If the name is empty, return a single dollar.
if len(name) == 0:
return '$'
# First try as a single variable; otherwise as an expression.
try:
v = getvariable(name)
if v is None:
v = eval(name, globals, locals)
except NameError:
return '$' + name
# See if the resulting value is one of the given special types.
try:
for objtype, objstr, objs in objlist:
if isinstance(v, objtype):
objs += [v]
return '$' + objstr + str(len(objs))
except:
pass
# No specific type, thus a normal value has to be substituted.
return substitutevar(v)
# This function tries to substitute the given name using
# the rules described in the description of function substitute.
def substituteexpr(expr, globals={}, locals={}):
try:
res = eval(expr, globals, locals)
v = substitutevar(res)
except:
# If the expr is undefined, return the original.
v = '$(' + expr + ')'
return str(v)
# Substitute a value.
def substitutevar(v):
out = ''
if isinstance(v, tuple) or isinstance(v, list) or isinstance(v, np.ndarray):
out = '['
first = True
for tmp in v:
if first:
first = False
else:
out += ','
out += substituteonevar(tmp)
out += ']'
else:
out = substituteonevar(v)
return out
def substituteonevar(v):
# A string needs to be enclosed in quotes.
# A vector value is enclosed in square brackets and separated by commas.
if isinstance(v, str):
return substitutestring(v)
# A numeric or boolean value is converted to a string.
# A vector value is enclosed in square brackets and separated by commas.
# Take care we have enough precision.
if isinstance(v, bool):
if v:
return 'T'
return 'F'
return str(v)
# Enclose a string in double quotes.
# If the string contains double quotes, enclose them in single quotes.
# E.g. ab"cd
# is returned as "ab"'"'"cd"
# which is according to the TaQL rules for strings.
def substitutestring(value):
out = '"'
for tmp in value:
if tmp == '"':
out += '"' + "'" + '"' + "'" + '"'
else:
out += tmp
return out + '"'
python-casacore-2.2.1/doc/ 0000775 0000000 0000000 00000000000 13203113151 0015346 5 ustar 00root root 0000000 0000000 python-casacore-2.2.1/doc/199.html 0000664 0000000 0000000 00001600054 13203113151 0016564 0 ustar 00root root 0000000 0000000
NOTE 199 – Table Query Language
NOTE 199 – Table Query Language
Ger van Diepen, ASTRON Dwingeloo
2016 Apr 4
Abstract
The Table Query Language (TaQL) is an SQL-like high level language to do operations likeselection, sort, and update on a casacore table. It is a very versatile language with full supportfor table columns containing array data. It has inherent support for masked arrays, units, andastronomical coordinates. It has a very rich set of functions (like cone search and array reduction)making it very suitable for astronomical applications. User deï¬ned functions can be added easily.It also has full support of grouping/aggregation and nested queries. An operation that can beexpressed in a single function is the matching of two sky catalogues. It can be used from C++, Python, and any shell.
The Table Query Language (TaQL, rhymes with bagel (though some people pronounce it as tackle)) is a
language for querying and manipulating data in Casacore tables. It makes it possible to get results or select
rows from an arbitrary table based on the contents of its columns and keywords. It supports arbitrary
complex expressions including units, extended regular expressions, and many functions. User deï¬ned
functions written in C++ or Python are supported. TaQL also makes sorting and column selection possible.
Furthermore TaQL has commands to update, add or delete rows and columns in a table and to create a new
table.
The ï¬rst sections of this document explain the syntax and show the options. The last sections give several
examples and show the interface to TaQL using Python or C++. The Python interface makes it possible to
embed Python variables and expressions in a TaQL command.
1.1 TaQL vs SQL
TaQL is modeled after SQL and contains a subset of SQL’s functionality. Some familiarity with SQL makes it
easier to understand the TaQL syntax. The most important features of TaQL different from SQL are:
The result of a SELECT is another table (either temporary or persistent). Usually this is a
so-called reference table, but it is also possible to make a deep copy and create a plain table. A reference table is a table that can be used as any other table, but does not contain data.
Instead it contains references to the rows and columns in the original table. Thus modifying data
in a reference table means that effectively the data in the original table are modiï¬ed.
A very rich set of mathematical and other functions.
Any operand can be a scalar or an N-dimensional array. Many reduce functions can be applied
to arrays.
Arrays can optionally be masked.
Full support of units and automatic conversion of units.
Support of various types of patterns/regular expressions and support of maximum string distance
(Levensthein (aka Edit) distance).
Speciï¬c operators and functions for cone searching (i.e., spatial searching with a search radius).
An advanced way of specifying intervals.
No support of indices, thus a linear table search is done. Because data are stored column-wise,
a linear search is usually very fast, even for very large tables.
Limited support for joins (only implicit joins on row number).
Many aggregate functions.
The COUNT command exists to count the occurrences of column values. Although it can still
be used, this command is obsolete now GROUPBY is fully supported.
The CALC command exists to calculate an arbitrary expression (including subqueries) on a table.
This can be useful to derive values from a table (e.g., the number of flags set in a measurement
set). It can even be used as a desk calculator.
TaQL can be used from languages with different conventions, for example the order of array
axes. Therefore it is possible to set the language style to be used.
The language can be extended by means of User Deï¬ned Functions, possibly implemented in
Python. Some standard UDFs exist to deal with MeasurementSets and to do measure conversions
(for directions, epochs, positions, and stokes).
TaQL has a keyword that makes it possible to time the various parts of a TaQL command.
2 TaQL Commands
2.1 Command summary
TaQL contains several commands. In the commands shown below the square brackets are not part of the
syntax, but indicate the optional parts of the commands.
show (or help)
  SHOW [type ...]
can be used to give some TaQL explanation or to show table information. A sole show command shows
the possible options. HELP is a synonym for SHOW.
selection
  SELECT [[DISTINCT] expression_list]
     [INTO table [AS options]]
     [FROM table_list]
     [WHERE expression]
     [GROUPBY expression_list]
     [HAVING expression]
     [ORDERBY [DISTINCT] sort_list]
     [LIMIT expression] [OFFSET expression]
     [GIVING table [AS options] | set]
     [DMINFO datamanagers]
It can be used to get an optionally sorted subset from a table. It can also be used to do a subquery
(see section 4.11 for more information on subqueries).
updating
  UPDATE table_list SET update_list [FROM table_list]
     [WHERE ...] [ORDERBY ...] [LIMIT ...] [OFFSET ...]
It can be used to update data in (a subset of) the ï¬rst table in the ï¬rst table list.
addition
  INSERT INTO table_list SET column=expr, column=expr, ...
 or
   INSERT INTO table_list [(column_list)] VALUES (expr_list)
 or
   INSERT INTO table_list [(column_list)] SELECT_command
It can be used to add and ï¬ll new rows in the ï¬rst table in the table list.
deletion
  DELETE FROM table_list
     [WHERE ...] [ORDERBY ...] [LIMIT ...] [OFFSET ...]
It can be used to delete some or all rows from the ï¬rst table in the table list.
counting
  COUNT [column_list] FROM table_list [WHERE ...]
It can be used to count occurrences of column values. Although the command can still be used, it is
basically obsolete because the same (and more) can be achieved with the GROUPBY clause and
aggregate functions in the SELECT command. Furthermore, usually GROUPBY is faster.
calculation
  CALC expression [FROM table_list]
It can be used to calculate an expression, in which columns in a table can be used.
table creation
  CREATE TABLE table [AS options]
     [column_spec]
     [LIMIT ...]
     [DMINFO datamanagers]
It can be used to create a new table with the given columns and number of rows. Optionally speciï¬c
table and data manager info can be given.
table structure modiï¬cation
  ALTER TABLE table
     [ADD COLUMN [column_spec] [DMINFO datamanagers]
     [RENAME COLUMN column_pair_list]
     [DROP COLUMN column_list]
     [SET KEYWORD key=value, key=value, ...]
     [COPY KEYWORD key=other, key=other, ...]
     [RENAME KEYWORD keyword_pair_list]
     [DROP KEYWORD keyword_list]
     [ADD ROW nrow]
It can be used to add, rename, and remove columns and keywords and to add rows. Multiple such
subcommands can be given, separated by white space.
TaQL can be used from different languages, in particular Python and Glish. Each has its own conventions
breaking down into three important categories:
0-based or 1-based indexing.
Fortran-order or C-order of arrays.
Inclusive or exclusive end in start:end ranges.
The user can set the style (convention) to be used by preceding a TaQL statement with
  USING STYLE value, value, ...
The possible (case-independent) values are:
BASE0 or BASE1 telling the indexing style.
ENDEXCL or ENDINCL telling the range style.
CORDER or FORTRANORDER telling the array style.
PYTHON which is equivalent to BASE0,ENDEXCL,CORDER
GLISH which is equivalent to BASE1,ENDINCL,FORTRANORDER
The following values are also possible and are described in the next subsections.
synonym=libname to deï¬ne a synonym for a user deï¬ned library.
TRACE or NOTRACE to (un)set tracing.
TIME or NOTIME to (un)set timing.
If multiple values are given for a category, the last one will be used. The default style used is GLISH, which is the
way TaQL always worked before this feature was introduced.
It is important to note that the interpretation of the axes numbers depends on the style being
used. e.g., when using glish style, axes numbers are 1-based and in Fortran order, thus axis 1 is
the most rapidly varying axis. When using python style, axis 0 is the most slowly varying axis.
Casacore arrays are in Fortran order, but TaQL maps it to the style being used. Thus when using python
style, the axes will be reversed (data will not be transposed). Note: unless said differently, all examplesin this document are done using the Python style.
The style feature has to be used with care. A given TaQL statement will behave differently if used with
another style.
2.2.1 UDF library synonyms
The style clause can also be used to deï¬ne synonyms for the library names of user deï¬ned functions. For
example:
  using style mscal=derivedmscal
deï¬nes the synonym mscal. Synonyms make it easier (i.e., less typing) to specify user deï¬ned functions.
Note that the synonym in the example above is automatically deï¬ned by TaQL as well as the synonym py for
pytaql.
2.2.2 Tracing
It is possible to get tracing output during the execution of a TaQL command by using the case-insensitive
value TRACE in the using style command.
2.2.3 Timing
It is possible to time a TaQL command by using the case-insensitive value TIME in the using style
command. For historical reasons it is also possible to use the the case-insensitive keyword TIME before or after
the optional style command.
Timing shows the total execution time and the times needed for various parts of the TaQL command on
stdout. For example:
time select distinct ANTENNA1,ANTENNA2
 from ~/3C343.MS where any(FLAG)’
Â
   Where          2.87 real    2.16 user    0.69 system
   Projection        0 real       0 user       0 system
   Distinct       0.18 real    0.16 user    0.03 system
  Total time      3.07 real    2.33 user    0.72 system
shows the time to do the where part (i.e., row selection on FLAG), projection (selection of columns), and
distinct (unique column values).
2.3 Reserved words
TaQL uses the following words as part of its language.
  ALL           AND      AS        ASC
   BETWEEN
   CALC          CREATETABLE
   DELETE        DESC     DISTINCT  DMINFO
   EXCEPT        EXISTS
   F             FALSE    FROM
   GIVING        GROUPBY  GROUPBYROLLUP
   HAVING
   IN            INCONE   INSERT    INTERSECT  INTO
   JOIN
   LIKE          LIMIT
   MINUS
   NODUPLICATES  NOT
   OFFSET        ON       OR        ORDERBY
   SAVETO        SELECT   SET       SUBTABLES
   T             TO       TOP       TRUE
   UNION         UNIQUE   UPDATE    USINGSTYLE
   VALUES
   WHERE
   XOR
These words are reserved. Note that the words in the TaQL vocabulary are case insensitive, thus the
lowercase (or any mixed case) versions are also reserved.
The reserved words cannot directly be used as column name, keyword name, or unit. However, a reserved
word can be used that way by escaping it with a backslash like \AS. When reading further, the meaning
of
     \IN  \in  IN [3mm,4mm]
    column unit IN    set
might become clear. It means: use unit in (inch) for column IN and test if it is in the given set.
Note this is unlike SQL where quotes have to be used to use a reserved word as a column name.
3 Selection from a table
The SELECT is the main TaQL command. It can be used to select a subset of rows and/or columns from a
table and to generate new columns based on expressions.
As explained above, the result of a selection is usually a reference table. This table can be used as any
other table, thus it is possible to do another selection on it or to update it (which updates the underlying
original table). It is, however, not possible to insert rows in a reference table or to delete rows from
it.
If the select column list contains expressions, it is not possible to generate a reference table. Instead a
normal plain table is generated (which can take some time if it contains large data arrays). It should be clear
that updating such a table does not update the original table.
The FROM clause can be omitted from the select. In that case no columns can be used in the selection,
but functions like rand and rowid make variable output possible. Clauses like ORDERBY can be given. The
GIVING (or INTO) might be useful to store the result in a table.
There is no explicit JOIN clause, but it is possible to equi-join tables on row number. Such tables must
have the same number of rows. One can also join , for example, the main table of a MeasurementSet with
a subtable like the ANTENNA table using a subquery. Joins are explained further in section
6.1.
3.1 SELECT command overview
The SELECT command consists of various clauses of which most are optional. The full command looks as
follows where the optional parts are shown in square brackets.
  SELECT [[DISTINCT] column_list]
     [INTO table [AS options]]
     [FROM table_list]
     [WHERE expression]
     [GROUPBY expression_list]
     [HAVING expression]
     [ORDERBY [DISTINCT] sort_list]
     [LIMIT expression] [OFFSET expression]
     [GIVING table [AS options] | set]
     [DMINFO datamanagers]
The clauses are executed in a somewhat different order.
FROM to deï¬ne the tables to be used.
WHERE to select the rows.
GROUPBY to group selected rows.
SELECT to ï¬ll select columns used in HAVING or ORDERBY
HAVING to select groups.
SELECT to ï¬ll the remaining select columns.
ORDERBY to sort the result.
LIMIT (or TOP) and OFFSET to ignore entries in the sorted result.
DMINFO to deï¬ne the data managers to be used if the result is
GIVING/INTO to store the ï¬nal result. stored in a plain table. See section 8.2 how to specify
them.
All clauses are explained in full detail in the subsequent sections.
3.1.1 Column/keyword lookup
Expressions in the various clauses will normally use column names to select, sort, or group a table. It is also
possible to use table keywords or column keywords by giving their names. Furthermore, it is possible to use a
column, created in the SELECT clause, in the HAVING and ORDERBY clauses. This can save time in both
specifying and executing the command, because a possibly complicated expression can be used to
create such a column. If such columns are used, that part of the SELECT is executed before
HAVING.
TaQL uses the following lookup scheme for column/keyword names.
If preceded by a shorthand (like in t0.DATA), the name is looked up in the corresponding table.
If not preceded by a shorthand, a name is ï¬rst looked up in the select columns. If not found, the
name is looked up in the ï¬rst table given in FROM.
A name is ï¬rst looked up as a column in a table. If not found, it is looked up as a table keyword.
Columns to be selected can be given as a comma-separated list with names of columns that have to be
selected from the tables in the table_list (see below). If no column_list is given, all columns of the
ï¬rst table will be selected. It results in a so-called reference table. Optionally a selected column
can be given another name in the reference table using AS name (where AS is optional). For
example:
  select TIME,ANTENNA1,ANTENNA2,DATA from 3C343.MS
   select TIME,ANTENNA1,ANTENNA2,MODEL_DATA AS DATA from 3C343.MS
It is possible to precede a column name with a table shorthand indicating with table in the FROM clause
has to be used. If not given, a column will be looked up in the ï¬rst table. Note that if equally named columns
from different tables are used, one has to get a new name, otherwise a ’duplicate name’ error will occur. For
example:
  select t0.DATA, t1.DATA as DATA1 from 3C343.MS t0, 3C343_1.MS t1
Apart from giving exact column names, it is also possible to use wildcards by means of a UNIX
ï¬lename-like pattern (like p/pattern/) or a regular expression (like f/regex/ for a full match or m/regex/
for a partial match). They can be suffixed with an i indicating case-insensitive matching. See
section 4.3.6 for a discussion of these constants. The operator ~ needs to be given before the
pattern or regex to indicate that columns have to be included. Thereafter operator !~ can be
used with another pattern or regex to remove columns. Such an excluding pattern or regex only
removes columns from the wildcarded columns before it until the latest non-wildcarded column.
A special pattern is * (which is the same as  p/*/). For example:
  select *, !~p/*_DATA/ from 3C343.MS
selects all columns except the ones ending in _DATA.
  select ~m/DATA/, !~p/*_DATA/ from 3C343.MS
selects columns with a name containing DATA except the ones ending in _DATA.
  select CORRECTED_DATA, *, !~p/*_DATA/ from 3C343.MS
 or
   select *, !~p/*_DATA/, CORRECTED_DATA from 3C343.MS
does select the CORRECTED_DATA column. Note it is not possible to change the name or data type of wildcarded columns.
It is also possible to use expressions in the column list to create new columns based on the contents of
other columns. When doing this, the resulting table is a plain table (because a reference table cannot contain
expressions). The new column can be given a name by giving AS name after the expression (where AS is
optional). If no name is given, a unique name like Col_1 is constructed. After the name a data type
string can be given for the new column. If no data type is given, the expression data type is
used.
  select max(ANTENNA1,ANTENNA2) AS ANTENNA from 3C343
   select means(DATA,1) from 3C343
Note that unit conversion can be (part of) an expression. For example:
  select TIME d AS TIMEH from my.ms
to store the time in unit d (days). Units are discussed in section 4.9.
It is possible to change the data type of a column by specifying a data type (see below) after the new
column name. Giving a data type (even if the same as the existing one) counts as an expression, thus results
in the generation of a plain table. For example:
  select MODEL_DATA AS DATA FCOMPLEX from 3C343.MS
Note that for subqueries the GIVING clause offers a better (faster) way of specifying a result expression.
It also makes it possible to use intervals.
Special aggregate functions (e.g., gmin) exist to calculate an aggregated value (minimum in this example)
per group of rows where the grouping is deï¬ned by the GROUPBY clause. The entire column
is a single group if no GROUPBY is given. Aggregation is discussed in more detail in section
5.
If a column_list is given and if all columns (and/or expressions) are scalars, the column_list can be
preceded by the word DISTINCT. It means that the result is made unique by removing the rows with
duplicate values in the columns of the column_list. Instead of DISTINCT the synonym NODUPLICATES or
UNIQUE can also be used. To ï¬nd duplicate values, some temporary sorting is done, but the original order of
the remaining rows is not changed. Note that support of this keyword is mainly done for SQL compliance. The same (and more) can be achieved
with the DISTINCT keyword in the ORDERBY clause with the difference that ORDERBY DISTINCT will
change the order. For full SQL compliance it is also possible to give the keyword ALL which is the opposite of DISTINCT, thus
all values are returned. This is the default. Because there is an ambiguity between the keyword ALL and
function ALL, the ï¬rst element of the column list cannot be an expression starting with a parenthesis if the
keyword ALL is used.
3.2.1 Masked array in column_list
If an expression in the column_list is a masked array, it is possible to create two columns from it: one for the
data, one for the mask. This can be done by combining them in parentheses like (DATA,MASK). A possible
data type given after the column names only applies to the data column, since the mask column always has
data type Bool. For example:
  select means(DATA[FLAG],0) as (MD,MM) C4 from in.ms giving out.tab
The select results in a masked array containing the means along axis 0. Both column MD and MM are ï¬lled
with the contents of the masked array. MD (with data type C4) contains the means over the ï¬rst axis of the
unmasked elements; MM contains the resulting mask.
[
AS options]]
3.3 INTO [table] [AS options]
This indicates that the ultimate result of the SELECT command should be written to a table (with the given
name). This table can be a reference table, a plain table, or a memory table.
The table argument gives the name of the resulting table. It can be omitted if a memory table is
created.
The options argument is optional and can be a single value or a list, enclosed in square brackets,
consisting of values and key=value. They can be used to specify the table and storage type. All keys and
values are case-insensitive.
TYPE=’value’
speciï¬es the table type. PLAIN = make a persistent table, thus a true copy of all selected rows/columns. SCRATCH = as plain, but only as a temporary table. MEMORY = as plain, but keep everything in memory. If TYPE is not given, a reference table is made if no expressions are given in the SELECT clause,
otherwise a plain table is made.
ENDIAN=’value’
speciï¬es the endianness BIG = big endian LITTLE = little endian LOCAL = native endianness of the machine being used AIPSRC = as deï¬ned in the .casarc ï¬le (which usually defaults to LOCAL) If ENDIAN is not given, it defaults to AIPSRC.
STORAGE=’value’
speciï¬es the storage type SEPFILE = store as separate ï¬les (the old Casacore table format) MULTIFILE = combine all storage manager ï¬les into a single ï¬le. MULTIHDF5 = as MULTIFILE, but use an HDF5 ï¬le instead of a regular ï¬le. DEFAULT = use SEPFILE (but might change in a future Casacore version), AIPSRC = as deï¬ned in the .casarc ï¬le (which usually defaults to DEFAULT) If STORAGE is not given, it defaults to AIPSRC.
BLOCKSIZE=n
speciï¬es the blocksize to use for MULTIFILE or MULTIHDF5.
OVERWRITE=F
tells that an existing table with the given name should not be overwritten. By
default TaQL will overwrite existing tables.
For backward compatibility, it is possible to specify an option directly without having to use ’key=value’.
MEMORY
to store the result in a memory table.
SCRATCH
to store the result in a scratch table, possibly on disk.
PLAIN
to store the result in a plain table.
PLAIN_BIG
to store the result in a plain table in big-endian format.
PLAIN_LITTLE
to store the result in a plain table in little-endian format.
PLAIN_LOCAL
to store the result in a plain table in native endian format.
The standard TaQL way to deï¬ne the output table is the GIVING clause. INTO is available for SQL
compliance.
If the INTO (or GIVING) clause is not given, the query result will be written into a memory table. In
this way queries done in a readonly directory will not fail if a result table cannot created. However, if the
result is expected to not ï¬t in memory (which will seldomly be the case), type SCRATCH should be used to
make it ï¬t.
If the result is stored in a plain table, it is possible to give detailed data manager info for
the result table using the DMINFO clause. See section 8.2 how the data manager info can be
speciï¬ed.
3.4 FROM table_list
The FROM part deï¬nes the tables used in the query. It is a comma-separated list of tables, each followed by
an optional shorthand (alias).
The full syntax is:
  FROM table1 [shorthand1], table2 [shorthand2], ...
Similar to SQL and OQL the shorthand can also be given using AS or IN. E.g.
  SELECT FROM mytable AS my, other IN ~user/othertable
Note that if using IN, the shorthand has to precede the table name. It can be seen as an iterator
variable.
The shorthand can be used in the query to qualify the table to be used for a column, for example
t0.DATA. The ï¬rst table in the list is the primary table which will be used if a column is not qualiï¬ed by a
shorthand. Often a query uses a single table in which case a shorthand is not needed. Multiple tables require
a shorthand and are useful when:
A keyword in another table is needed.
Columns from multiple tables are used (an implicit join). In such a case the tables must have the same
number of rows. For example, a regression test could be done like:
  SELECT FROM test.MS t1, result.MS t2
     WHERE not all(near(t1.DATA, t2.DATA))
If the table is normal table with a fully alphanumeric name, the shorthand defaults to that name. In practice a
shorthand is always needed if multiple tables are used.
The FROM clause can be omitted, in which case the input is a virtual table with no columns. The
number of rows in it is deï¬ned by the LIMIT and OFFSET value; it defaults to 1 row. It makes it possible to
select column-independent expressions in the SELECT command. Note that these expressions do not need to
be constant. For example
  SELECT rowid() LIMIT 31
creates a temporary table with column Col_1 and 31 rows containing the values 0..30.
A table can be given in a variety of ways.
A persistent table can be used by giving its name which can contain path speciï¬cation and
environment variables or the UNIX ~ notation. If the tablename contains a special character,
the character can be escaped with a backslash or the table name can be enclosed in single or
double quotes.
A table name can be taken from a keyword in a previously speciï¬ed table. This can be useful in a
subquery. The syntax for this is the same as that for specifying keywords in an expression.
E.g.
  SELECT FROM mytable tab
     WHERE col1 IN [SELECT subcol FROM tab.col2::key]
In this example key is a table keyword of column col2 in table mytable (note that tab is the
shorthand for mytable and could be left out). It can also be used for another table in the main query. E.g.
  SELECT FROM mytable, ::key subtab
     WHERE col1 > subtab.key1
In this example the keyword key1 is taken from the subtable given by the table keyword key in the
main table. If a keyword is used as the table name, the keyword is searched in one of the tables previously given.
The search starts at the current query level and proceeds outwards (i.e., up to the main query level). If
a shorthand is given, only tables with that shorthand are taken into account. If no shorthand is given,
only primary tables are taken into account.
Opening a subtable using a path name like my.ms/ANTENNA will fail if my.ms is a reference table instead
of the original table. Therefore the path of a subtable should be given using colons instead of slashes
like my.ms::ANTENNA which is a slight extension of specifying table names in the previous bullet.
In this way a subtable can always be found.
Similar to OQL it is possible to use a nested query command in the FROM clause. This is a normal
query command enclosed in square brackets or parentheses. Besides the SELECT command the
COUNT and CREATE TABLE command can also be used. The table created can thereafter be used in
the rest of the query command by using the shorthand given to that table. It can also be used in the
remainder of the table_list, thus using it as a backreference. Such backreferencing can be useful to avoid
multiple equal subqueries. E.g.
  select from MS,
    [select from MS where sumsqr(UVW[1:2]) < 625]
     as TIMESEL
    where TIME in [select distinct TIME from TIMESEL]
     &&  any([ANTENNA1,ANTENNA2] in
       [select from TIMESEL giving
         [iif(UVW[3] < 0, ANTENNA1, ANTENNA2)]])
is a command to ï¬nd shadowed antennas for the VLA. Without the query in the FROM command the
subqueries in the remainder of the command would have been more complex. Furthermore, it would
have been necessary to execute that select twice. The command above is quite complex and cannot be fully understood before reading the rest of this
note. Note, however, that the command uses the shorthand TIMESEL to be able to use the temporary
table in the subqueries.
Normally only persistent tables (i.e., tables on disk) can be used. However, it is also possible to use
transient tables in a TaQL command given in Python, Glish, or C++. This is done by passing one or
more table objects to the function executing the TaQL command. In the TaQL command a $-sign
followed by a sequence number has to be given to indicate the correct object containing the transient
table. E.g., if two table objects are passed $1 indicates the ï¬rst table, while $2 indicates the second
one.
It is possible to use a concatenation of tables with the same description by giving a list of tables
enclosed in square brackets. In this way it is, for example, possible to do a query on the
combined parts of a MeasurementSet partitioned in time. Each table in the list can be
speciï¬ed in one of the ways mentioned in this section, including another table concatenation.
For example:
  SELECT FROM [ms.part1, ms.part2, ms.part3] WHERE ...
does a query on the three parts of an MS which are seen as a single table. It is possible to use glob ï¬lename patterns in such a list. For example
  SELECT FROM [ms.part*] WHERE ...
is the same as the example above if no other ï¬les with such a name exist. An error is given if no table
is found matching the pattern.
Subtables of the concatenated tables can be concatenated as well. Alternatively, they can be assumed
to be the same for all tables meaning that the subtable of the concatenation is the subtable of the
ï¬rst table. For example, when partitioning a MeasurementSet in time, the ANTENNA
subtable is the same for all parts, while the POINTING and SYSCAL subtables depend on
time, thus have to be concatenated as well. Concatenation of subtables can be achieved
by giving them as a comma-separated list of names after the SUBTABLES keyword. For
example:
  SELECT FROM [ms.part1, ms.part2 SUBTABLES SYSCAL,POINTING]
Usually the result of a TaQL query references the table given in the FROM. In this example the
FROM table is the concatenation, which is only known during the query. In such a case the
concatenation must be made persistent, which can be done by using a GIVING (or INTO) inside
the concatenation speciï¬cation. Only the table name can be given, because the persistent
concatenation only keeps the original table names; it does not make a copy of all data.
For example:
  SELECT FROM [ms.part1, ms.part2 GIVING ms.conc]
      WHERE ANTENNA1 != ANTENAA2 GIVING ms.cross
selects the cross-correlation baselines from the concatenation. Note the two GIVING commands. The
ï¬rst one makes the concatenation persistent, the second one is the query result of the query ms.cross. It
references the matching rows in the persistent concatenation ms.conc which in its turn references the
original parts.
3.5 WHERE expression
It deï¬nes the selection expression which must have a boolean scalar result. A row in the primary table is
selected if the expression is true for the values in that row. The syntax of the expression is explained in a
section 4.
3.6 GROUPBY group_list
It deï¬nes how rows have to be grouped. Usually a result per group will be calculated using aggregate
functions. A group consists of all rows for which the columns (or expressions) given in the group_list have the
same value. The (aggregate) expressions in the SELECT clause are calculated for the entire group. In this
way one can get, for example, the mean XX amplitude and the number of time slots per baseline
like:
  SELECT ANTENNA1,ANTENNA2,GMEAN(AMPLITUDE(DATA[,0])),GCOUNT()
          FROM my.ms GROUPBY ANTENNA1,ANTENNA2
It results in a table containing nbaseline rows with in each row the antenna ids, mean amplitude, and
number of rows. If no aggregate function is used for a column, the value of the last row in the group is used. Note that in this
example ANTENNA1 and ANTENNA2 are the same for the entire group. However, if TIME was also
selected, only the last time would be part of the result. Note that each expression in the group_list has to result in a scalar value of type bool, integer, double, date,
or string. Aggregate functions are discussed in more detail in section 5.
3.7 HAVING expression
This clause can be used to select speciï¬c groups. Only the groups (deï¬ned by GROUPBY) are selected for
which the HAVING expression is true. Note that HAVING can be given without GROUPBY, although that will hardly ever be useful. If no
GROUPBY is given, but the SELECT statement contains an aggregate function, the result is a single
group. HAVING cannot be used if neither GROUPBY nor SELECT aggregate functions are used.
It is discussed in more detail in section 5.
3.8 ORDERBY sort_list
It deï¬nes the order in which the result of the selection has to be sorted. The sort_list is a comma separated
list of expressions. It operates on the output of the SELECT, thus after a possible GROUPBY and HAVING
clause are executed. The sort_list can be preceded by the word ASC or DESC indicating if the given expressions are by default
sorted in ascending or descending order (default is ASC). Each expression in the sort_list can
optionally be followed by ASC or DESC to override the default order for that particular sort key.
To be compliant with SQL whitespace can be used between the words ORDER and BY.
The word ORDERBY can optionally be followed by DISTINCT which means that only the ï¬rst row of
multiple rows with equal sort keys is kept in the result. To be compliant with SQL dialects the word
UNIQUE or NODUPLICATES can be used instead of DISTINCT.
An expression can be a scalar column or a single element from an array column. In these cases some
optimization is performed by reading the entire column directly. It can also be an arbitrarily complex expression with exactly the same syntax rules as the expressions in
the WHERE clause. The resulting data type of the expression must be a standard scalar one,
thus it cannot be a Regex or DateTime (see below for a discussion of the available data types).
E.g.
  ORDERBY col1, col2, col3
   ORDERBY DESC col1, col2 ASC, col3
   ORDERBY NODUPLICATES uvw[1] DESC
   ORDERBY square(uvw[1]) + square(uvw[2])
   ORDERBY datetime(col)       # incorrect data type
   ORDERBY mjd(datetime(col))  # is correct
3.9 LIMIT/OFFSET expression
It indicates which of the matching and sorted rows should be selected. If not given, all of them are selected.
The word TOP can also be used instead of LIMIT. LIMIT and OFFSET are applied after ORDERBY and SELECT DISTINCT, so they are particularly useful in
combination with those clauses to select, for example, the highest 10 values.
It can be given in two ways:
In the semi-standard SQL way using LIMIT N to select N rows and/or OFFSET M to skip the ï¬rst
M rows. Similar to Python, N and M can be negative meaning they are counted from the end.
E.g., LIMIT -1 means all rows but the last.
As a Python-style range using LIMIT start:end:incr, where the end is exclusive. Start defaults
to 0, end to the number of rows, and incr to 1. As above, start and end can be negative to count
from the end. The increment must be positive.
For example:
  SELECT FROM my.tab ORDERBY DISTINCT TIME LIMIT 2 OFFSET 10
   SELECT FROM my.tab ORDERBY DISTINCT TIME LIMIT 10:12
sorts uniquely by time, skips the ï¬rst 10 rows, and selects the next two rows.
  SELECT FROM my.tab LIMIT ::100
selects every 100-th row.
[
AS options] — set]
3.10 GIVING [table] [AS options] — set
It indicates that the ultimate result of the SELECT command should be written to a table (with the given
name). Another (more SQL compliant) way to deï¬ne the output table is the INTO clause. See INTO for a more
detailed description including the possible types.
It is also possible to specify a set in the GIVING clause instead of a table name. This is very useful if the
result of a subquery is used in the main query. Such a set can contain multiple elements Each element can be
a single value, range and/or interval as long as all elements have the same data type. The parts of each
element have to be expressions resulting in a scalar.
In the main query and in a query in the FROM command the GIVING clause can only result in a table
and not in a set. To be compliant with SQL dialects, the word SAVETO can be used instead of GIVING. Whitespace can be
given between SAVE and TO.
4 Expressions
An expression is the basic building block of TaQL. They are similar to expressions in other languages. An
expression is formed by applying an operator or a function to operands which can be a table column or
keyword, a constant, or a subexpression. An operand can be a scalar value or an array or set. The next
subsections discuss them in detail.
An expression can be used in several places:
In the WHERE and HAVING clause where the result must be a boolean scalar value. It tells if
a table row or group will be selected.
As a key in the GROUPBY clause where the result must be a scalar value (numeric, bool, or
string).
As a sort key in the ORDERBY clause where the result must be a scalar value (numeric, bool,
or string)
As an element in the set in the GIVING clause. It must be a scalar value of any type except
regex.
As a scalar or array value in the INSERT and UPDATE command.
As a column expression in the column-list part of the SELECT command. The result can be a
scalar or array value.
As a scalar or array value in the CALC command.
As a scalar or array value in various ALTER TABLE subcommands
The expression in the clause can be as complex as one likes using arithmetic, comparison, and logical
operators. Parentheses can be used to group subexpressions. The operands in an expression can be table columns, table keywords, constants, units, functions, sets and
intervals, and subqueries. The index operator can be used to take a single element or a subsection from an array expression.
For example,
  column1 > 10
   column1 + arraycolumn[index] >= min (column2, column3)
   column1 IN [expr1 =:< expr2]
The last example shows a set with a continuous interval.
4.1 Data Types
Internally TaQL uses the following data types:
Bool
logical values (true/false (case-insensitive) or T/F)
Integer
integer numbers up to 64 bits
Double
64 bit floating point numbers including times/positions
Complex
128 bit complex numbers
String
string values on which operator + can be used (concatenation).
Regex
regular expressions can be used for string matching (see section 4.2). Maximum string
distances can also be used in a way similar to regular expressions.
DateTime
representing a date/time. There are several functions acting on a date/time. Operator
+ and - can be used on them.
Scalars and arbitrarily shaped arrays of these data types can be used. However, arrays of Regex are not possible.
If an operand or function argument with a non-matching data type is used, TaQL will do the following
automatic conversions: - from Integer to Double or Complex. - from Double to Complex. - from String or Double to DateTime.
In this document some special data types are used when describing the functions. - Real means Integer or Double. - Numeric means Integer, Double, or Complex. - DNumeric means Double or Complex.
TaQL supports any possible data type of a table column or keyword. In some commands (column list and
CREATE TABLE) columns are created where it is possible to specify the data type of a column. The
following case-insensitive values can be used to specify a type:
  B          BOOL       BOOLEAN
   U1         UCHAR      BYTE
   I2         SHORT      SMALLINT
   U2   UI2   USHORT     USMALLINT
   I4         INT        INTEGER
   U4   UI4   UINT       UINTEGER
   R4   FLT   FLOAT
   R8   DBL   DOUBLE
   C4   FC    FCOMPLEX   COMPLEX
   C8   DC    DCOMPLEX
   S          STRING
   TIME       DATE       EPOCH
The TIME type is a special data type. It means that the column gets data type DOUBLE and that
a MEASINFO record will be deï¬ned in the column keywords to designate the column as an
epoch.
4.2 Regular Expressions and String Distances
TaQL supports the use of extended regular expressions and string distances. They can be speciï¬ed in
various ways as discussed in section 4.3.6. There are three basic types of regular expressions.
An SQL-style pattern is quite simple. It has 2 special characters. The underscore (_) means a
single arbitrary character and the percent (%) means zero or more arbitrary characters.
Special characters can be escaped with a backslash to retain their normal meaning. For
example:
  3c\_%
matches 3c_ and 3c_xx, but not 3caxx.
A UNIX-style pattern, as often used for wildcarded ï¬le names, is more powerful than the SQL-style
pattern. It has a few special characters that can be escaped with a backslash.
The question mark (?) means a single arbitrary character.
The asterisk (*) means zero or more arbitrary characters. For example: 3c_* does the same
as the SQL-style pattern above.
Square brackets indicate a bracket expression (character choice). For example: [ab] matches a and
b, but not c. A few special characters can be used in a bracket expression:
A leading ^ or ! means negation. Thus [!ab] matches every character except a and b.
A minus sign indicates a range. For example [0-9] matches a digit or [a-z] matches
a lowercase letter. If a minus sign cannot be interpreted as a range, it is a literal minus
sign like in [-ab] or the second minus sign in [a-z-A].
Posix character classes [:xx:] where xx can be: - alpha matching any letter - lower matching any lowercase letter - upper matching any uppercase letter - alnum matching any digit or letter - digit matching any digit - xdigit matching any hexadecimal digit (0-9a-fA-F) - space matching any whitespace character - print matching any printable character (alnum, punct, space) - punct matching any non-alnum visible character (.,!? etc.) - graph matching any visible printable character (alnum, punct) - cntrl matching any control character. For example [_[:isalpha:]][_[:isalnum:]]* to match variable names.
A bracket expression cannot be empty, thus if ] is the ï¬rst character in the bracket
expression, it is interpreted literally. Note that is also true if it is the ï¬rst character
after the negation character.
A backslash in a character class is always interpreted literally, thus special characters
cannot be escaped. However, as shown above they can always be placed such that they
are interpreted literally.
Braces can be used for a choice between (possible empty) multi-character strings separated by
commas. Escape a comma or brace with a backslash to treat it literally. For example:
*.{h,hpp,c,cc,cpp} It is fully nestable, thus choice strings can be patterns. For example: *.{[hc]{,pp},c} does the same as the example above. Note that the inner choice is between an empty string and
pp.
An awk/egrep-like extended regular expression is most powerful. A full explanation can be found on
Wikipedia. Here only a summary of its special characters is given. They can be escaped using
backslashes.
. matches any character.
^ matches beginning of string.
$ matches end of string.
Square brackets for a bracket expression. It is the same as described above with the
exception that ! cannot be used as negation character.
* matches zero or more occurrences of previous character or subexpression.
+ matches one or more occurrences.
? matches zero or one occurrence.
{ and } for an interval giving minimum and maximum number of occurrences. For example:
[a-z]{3,5} matches lowercase string with a minimum of 3 and maximum of 5 characters.
[a-z]{3} matches exactly 3 characters. [a-z]{3,} matches at least 3 characters. [a-z]{,5} matches at most 5 characters.
| matches left or right substring
( and ) to form subexpressions for operators like *.
\1 till \9 mean backreference to a subexpression (ï¬rst one is \1). A string part matches if
it is equal to the string part matching that subexpression. e.g., (a*)x\1 matches x, axa,
aaxaa, etc., but not axaa nor aaxa.
For example:
  .*\.(h|hpp|c|cc|cpp)
   .*\.[hc](pp)?|cc
do the same as the pattern examples above.
Furthermore it is possible to specify maximum string distances (known as Levensthein or Edit distance). It is
explained in section 4.3.6.
  column ~ d/string/ibnn
4.3 Constants
Scalar constants of the various data types can be formed in a way similar to Python and Glish. Array
constants can be formed from scalar constants.
4.3.1 Bool
A Bool constant is the value T or F (both in uppercase) or the value true or false (any case).
4.3.2 Integer
An integer constant is a numeric value without decimal point or exponent. It can also be given as a
hexadecimal value like 0xffff.
4.3.3 Double (and time/position)
A floating-point constant is given with a decimal point and/or exponent. ’E’ or ’e’ can be used to
specify the exponent. An integer number followed by a unit is also regarded as a double constant.
Another way to deï¬ne a Double constant is by means of a Time or Position. Such a constant is always
converted to radians. It can be given in several ways:
An integer or floating-point number immediately followed by a simple unit (thus without
whitespace). e.g., 12.43deg Some valid units are deg, arcmin, arcsec (or as), rad. The units can be scaled by preceding them
with a letter (e.g., mrad is millirad).
A time/position in HMS format. Seconds can be left out. e.g., 12h34m34.5 or 8h32m
A position in DMS format. Seconds can be left out. e.g., 12d34m34.5 or 8d0m
A position as DMS in dot format. Note that all parts must be present. e.g., 12.34.34.5 or
8.0.34.5
4.3.4 Complex
The imaginary part of a Complex constant is formed by an Integer or Double constant immediately followed
by a lowercase i or j. A full Complex constant is formed by adding another Integer or Double constant as the
real part. E.g.
  1.5 + 2j
   2i+1.5            is identical
  Â
Note that a full Complex constant has to be enclosed in parentheses if, say, a multiplication is performed on
it. E.g.
  2 * (1.5+2i)
  Â
4.3.5 String
A String constant has to be enclosed in †or ’ and can be concatenated (as in C++). E.g.
  "this is a string constant"
   ’this is a string constant containing a "’
   "ab’cd"’ef"gh’
       which results in:   ab’cdef"gh
  Â
4.3.6 Regular expression and String distance
A regular expression constant can be given directly or using a function.
An SQL-style pattern can be given directly as a string constant preceded by operator LIKE or
NOT LIKE.
A pattern or regular expression can be given like x/expr/q preceded by operator ~ or !~. Instead of a
slash, the characters % and # can also be used as delimiter, as long as the same delimiter is used on
both sides. The delimiter can not be part of the expression (not even escaped with a backslash).
The x denotes the type: - p means a pattern matching the full string. - f means a regular expression matching the full string. - m means a regular expression matching part of the string (a la Perl). The q denotes optional qualiï¬ers. Currently only i is supported meaning a case-insensitive match. For
example:
  name~p/3[cC]*/
   name ~ p%3c*%i
   lower(name) ~ p%3c*%
   name ~ m/^3c/i
   name ~ f/3c.*/i
   filename !~ p#/usr/*.{h,cc}#
All examples but the last one do the same: matching a name starting with 3c or 3C. The last example shows a glob-style pattern to ï¬nd ï¬les on /usr not ending in .h or .cc.
Apart from these Perl-like speciï¬cations, a regular expression can also be formed by applying a function
to a string constant. The operator = or != has to be applied to it.
Function sqlpattern treats its argument as an SQL-style pattern. For example:
  lower(name) LIKE ’3c%’
   lower(name) = sqlpattern(’3c%’)
do the same.
Function pattern treats its argument as a UNIX-style pattern.
Function regex treats its argument as a full regular expression.
Case-insensitive matching can only be done as shown in the example above by downcasing the string to
be matched. Please note that these functions are not limited to constants. They can also be used to form regular
expressions from variables.
A maximum string distance constant can be speciï¬ed in a similar way. Such a distance is known as the
Levensthein or Edit distance. It is a measure of the similarity of strings by counting the minimum number of
edits (deletions, insertions, substitutions, and swaps of adjacent characters) that need to be done to make the
strings equal.
  column ~ d/string/ibnn
This tests if the strings in the given column are within the maximum distance of the string given in the
constant. The following qualiï¬ers can be given (in any order):
i means a case insensitive test.
b means that blanks in the strings are ignored.
nn is an integer value giving the maximum distance. If not given it defaults to 1 + len(string)/ 3.
4.3.7 Date/time
DateTime constant can be formed in 2 ways:
From a String constant using the datetime function. In this way all possible formats as explained in
class MVTime are supported. E.g.
  datetime ("11-Dec-1972")
A more convenient way is to specify it directly. Since this makes use of the delimiters space, - or /, it
conflicts with the expression grammar as such. However, possible conflicts can be solved by using
whitespace in an expression and it is believed that in practice the convenience surpasses the possible
conflicts. A large subset of the MVTime formats is supported. A DateTime has to be speciï¬ed as date/time or
date-time, where the time part (including the space, -, or / delimiter) is optional. The possible date
formats are: - YYYY/MM/DD, YYYY-MM-DD, or DD-MM-YYYY where DD and MM must be 2 digits and
YYYY 4 digits. - DD-MMMMMMMM-YY where the - is optional and MMMMMMM is the case-insensitive name of
the month (at least 3 letters). DD can be 1 or 2 digits and YY 1 to 4 digits. 2000 is added if
YY50 and 1900 is added if
50YY100.
If MM12,
YYYY will be incremented accordingly. The general time format in a DateTime constant is: - hh:mm:ss.s where the delimiter h or H can be used for the ï¬rst colon and m or M for the second. Trailing parts
can be omitted. E.g.
  10-02-1997
   10-February-97
   10feb97
   1997/02/10         are all identical
Â
   1May96/3:          : (or h) is mandatory
   1May96-3:0
   1May96 3:0:0
   1May96-3h          h (or :) is mandatory
   1May96 3H0
   1May96/3h0M
   1May96/3hm0.0
A DateTime constant with the current date/time can be made by using the function datetime without
arguments.
4.3.8 Arrays
N-dimensional arrays of all data types can be created with the exception of regular expressions.
It is possible to form a 1-dimensional array from a constant bounded discrete set. When needed such a set is
automatically transformed to an array. E.g.
  [0:10]
   [’str1’, ’str2’, ’str3’]
   ’str’ + [’1’, ’2’, ’3’]
The ï¬rst example results in an integer array of 10 elements with values 0..9. The others result in a string
array of 3 elements. The second version already shows that strings can be concatenated (as explained further
on).
A multi-dimensional array can be formed by giving a set of arrays. A nested list resembles the numpy
way. For example:
  [[1,2,3],[4,5,6]]
results in a 2-dim array. However, it is also possible to use arrays created in other ways such as arrays in a
column or arrays created with the array function described below. For example:
  [[[1,2,3],[4,5,6]], array([10:13],2,3)]
results in a 3-dim array.
Furthermore it is possible to use the array function to create an array of any shape. The values are given
in the ï¬rst argument as a scalar, set, or another array. The shape is given in the latter arguments as scalars
or as a set. The array is initialized to the values given which are wrapped if the array has more
elements.
  array([1:11],10,4)
   array([1:11], [10,4])
   array(F,shape(DATA))
The ï¬rst examples create an array with shape [10,4] containing the values 1..10 in each line.
The latter results in a boolean array having the same shape as the DATA array and ï¬lled with
False.
4.3.9 Masked Arrays
An array can have an optional mask. Similar to numpy’s masked array, a mask value True means that the
value is masked off, thus not taken into account in reduce functions like calculating the mean.
Note that this deï¬nition is the same as the FLAG column in a MeasurementSet, but is different from a mask
in a Casacore Image where True means good and False means bad.
All operations on arrays will take the possible mask into account. Reduce functions like median only use
the unmasked array elements. Furthermore, partial reduce functions like medians will set an
output mask element to True if the corresponding input array part has no unmasked elements.
Operators like + and functions like cos operate on all array elements. The mask in the resulting array is
the logical OR of the input masks. Of course, the result has no mask if no input array has a
mask.
A masked array is created by applying a boolean array to an array using the square brackets operator.
Both arrays must have the same shape. For example:
  DATA[FLAG]
   DATA[DATA > 3*median(abs(DATA))]
The ï¬rst example applies the FLAG column in a MeasurementSet to the DATA column. The second
example masks off high DATA values.
The functions arraydata, arraymask, and flatten can be used to get the array data or mask. The last
one flattens the array to a vector while removing all masked elements.
The TaQL commands putting values into a table accept two columns (in parentheses) for a masked array.
This is described in more detail in the appropriate sections. For example:
  select means(DATA[FLAG],0) as (MD,MM) from in.ms giving out.tab
to write the data averaged over the ï¬rst axis (frequency channel)into column MD. Only the unflagged data
points are taken into account. The output contains the resulting flags in column MM; a flag is set to True if
all channels were flagged.
4.3.10 Null Arrays
A cell in a table column containing variable shaped arrays, can be empty. Such a cell does not contain an
array and is represented in TaQL as a null array. Note it is different from a cell containing an empty array,
which is an array without values.
Null arrays can be used with any operator and in any function. If one of the operands or function
arguments is a null array, the result will be a null array; only array functions reducing to a scalar (such as
sum and mean) give a valid value (usually 0).
The UPDATE and INSERT commands will ignore a null array result; no value is written in that
row.
4.4 Table Columns
A table column can be used in a query by giving its name in the expression, possibly qualiï¬ed with a table
shorthand name. A column can contain a scalar or an array value of any data type supported by the table
system. It will be mapped to the available TaQL data types. If the column keywords deï¬ne a unit for the
column, the unit will be used by TaQL.
The name of a column can contain alphanumeric characters and underscores. It should start with an
alphabetic character or underscore. A column name is case-sensitive. It is possible to use other characters in the name by escaping them with a backslash. e.g., DATE\-OBS.
In the same way a numeric character can be used as the ï¬rst character of the column name. e.g., \1stDay.
A reserved word cannot be used directly as a column name. It can, however, be used by escaping it with a
backslash. e.g., \IN. Note that in programming languages like C++ and Python a backslash itself has to be escaped by another backslash. e.g., in Python:
tab.query(’DATE\\-OBS10MAR1996’).
If a column contains a record, one has to specify a ï¬eld in it using the dot operator; e.g., col.fld means
use ï¬eld fld in the column. It is fully recursive, so col.fld.subfld can be used if ï¬eld fld is a record in
itself. Alas records in columns are not really supported yet. One can specify ï¬elds, but thereafter an error message
will be given.
4.4.1 Referring to SELECT columns
Usually a column used in an expression will be a column in one of the tables speciï¬ed in the FROM clause.
However, it is possible to use a column created in the SELECT clause, in expressions given in the HAVING
or ORDERBY clause. In fact, a column name not preceded by a table shorthand, is ï¬rst looked up in the
SELECT columns and thereafter in the ï¬rst FROM table.
It can be advantageous to use a SELECT column if that column is an expression; it saves both typing and
execution time. because that expression is executed only once.
4.5 Table Keywords
It is possible to use table or column keywords, which can have a scalar or an array value or a record, possibly
nested. A table keyword has to be speciï¬ed as ::key. In an expression the :: part can be omitted if
there is no column with the same name. A column keyword has to be speciï¬ed as column::key.
Note that the :: syntax is chosen, because it is similar to the scope operator in C++. As explained in the FROM clause, keywords in the primary table and in other tables can be used. If
used from another table, it has to be qualiï¬ed with the (shorthand) name of the table. E.g.,
sh.key or sh.::key takes table keyword key from the table with the shorthand name sh.
If a keyword value is a record, it is possible to use a ï¬eld in it using the dot operator. e.g., ::key.fld to
use ï¬eld fld. It is fully recursive, so if the ï¬eld is a record in itself, a subï¬eld can be used like
col::key.fld.subfld
A keyword can be used in any expression. It is evaluated immediately and transformed to a constant
value.
4.6 Operators
TaQL has a fair amount of operators which have the same meaning as their C and Python counterparts. The
operator precedence order is:
  **
   !  ~  +  -       (unary operators)
   *  /  // %
   +  -
   &
   ^
   |
   == != >  >= <  <=  ~= !~= IN INCONE BETWEEN EXISTS LIKE  ~  !~
   &&
   ||
Operator names are case-insensitive. For SQL compliancy some operators have a synonym.
  ==     =
   !=     <>
   &&     AND
   ||     OR
   !      NOT
   ^      XOR
All operators can be used for scalars and arrays and a mix of them. Note that arrays of regular expressions
cannot be used.
The following table shows all available operators and the data types that can be used with
them.
Operator
Data Type
Description
**
numeric
power. It is right associative, thus 2**1**2 results in 2.
*
numeric
multiplication
/
numeric
non-truncated division, thus 1/2 results in 0.5
//
real
truncated division (a la Python) resulting in an integer, thus
1./2. results in 0
%
real
modulo; 3.5%1.2 results in 1.1; -5%3 results in -2
+
no bool
addition. If a date is used, only a real (converted to unit day)
can be added to it. String addition means concatenation.
-
numeric,date
subtraction. Substracting a date from a date results in a real
(with unit day). Subtracting a real (converted to unit day)
from a date results in a date.
&
integer
bitwise and
|
integer
bitwise or
^, XOR
integer
bitwise xor
,
all
comparison for equal. The norm is used when comparing
complex numbers.
no bool
comparison for greater
no bool
comparison for greater or equal
no bool
comparison for less
no bool
comparison for less or equal
,
all
comparison for not equal
~=
numeric
shorthand for the NEAR function with a tolerance of 1e-5
test if a subquery ï¬nds at least N matching rows. The value
for N is taken from its LIMIT clause; if LIMIT is not given it
defaults to 1. The subquery loop stops as soon as N matching
rows are found. E.g. EXISTS(select from ::ANTENNAwhere NAME=’’somename’’ LIMIT 2) results in true if at
least 2 matching rows in the ANTENNA table were found.
NOT EXISTS
negation of EXISTS
4.7 Sets and intervals
As in SQL the operator IN can be used to do a selection based on a set. E.g.
  SELECT FROM table WHERE column IN [1,2,3]
The result of operator IN is true if the column value matches one of the values in the set. A set can contain
any data type except a regex.
This example shows that (in its simplest form) a set consists of one or more values (which can be
arbitrary expressions) separated by commas and enclosed in square brackets. The elements in a set have to be
scalars and their data types have to be the same or convertible to a common data type. The square brackets
can be left out if the set consists of only one element. For SQL compliance parentheses can be used instead of
square brackets.
An array is also a set, so IN can also be used on an array like:
  SELECT FROM table WHERE column IN expr1
where expr1 is the array result of some expression. It is also possible to use a scalar as the
righthand of operator IN. So if expr1 is a scalar, operator IN gives the same result as operator
==.
The lefthand operand of the IN operator can also be an array or set. In that case the result is a
boolean array telling for each element in the lefthand operand if it is found in the righthand
operand.
An element in a set can be more complicated than a single value. It can deï¬ne multiple values and
intervals. The possible forms of a set element are:
A single value as shown in the example above.
start:end:incr. This is similar to the way an array index is speciï¬ed. Incr defaults to 1. End defaults
to an open end (i.e., no upper bound) and results in an unbounded set. Start and end can be a real or a
datetime. Incr has to be a real. Some examples:
  1:10         means 1,2,...,9  (also 10 when using glish style)
   1:10:2       means 1,3,5,7,9
   1::2         means all odd numbers
   1:           means all positive integer numbers
   18Aug97::2   means every other day from 18Aug97 on
These examples show constants only, but start, end, and incr can be any expression.
Note that :: used here can conflict with the :: in the keywords. e.g., a::b is scanned as a keyword
speciï¬cation. If the intention is start::incr, whitespace should be used as in a: :b. In practice this
conflict will hardly ever occur.
Continuous intervals can be speciï¬ed for data type real, string, and datetime. The speciï¬cation of an
interval resembles the mathematical notation 1<x<5, where x is replaced by :. An open interval side is
indicated by <, while a closed interval side is indicated by =. Another way to specify intervals is using curly and/or angle brackets. A curly bracket is a closed side,
the angle bracket is an open side. The following examples show how bounded and half-bounded,
(half-)open and closed intervals can be speciï¬ed.
  1=:=5   {1,5}     means 1<=x<=5   bounded closed
   1<:<5   <1,5>     means 1<x<5     bounded open
   1=:<5   {1,5>     means 1<=x<5    bounded right-open
   1<:=5   <1,5}     means 1<x<=5    bounded left-open
   1=:  {1,}  {1,>   means 1<=x      left-bounded closed
   1<:  <1,}  <1,>   means 1<x       left-bounded open
   :=5  {,5}  <,5}   means x<=5      right-bounded closed
   :<5  {,5>  <,5>   means x<5       right-bounded open
It is very important to note that the 2nd form of set speciï¬cation results in discrete values, while the 3rd form
results in a continuous interval.
Each element in a set can have its own form, i.e., one element can be a single value while another can be
an interval. If a set consists of single or bounded discrete start:end:incr values only, the set will be
expanded to an array. This makes it possible for array operators and functions (like mean) to be applied to
such sets. E.g.
  WHERE column > mean([10,30:100:5])
Another form of constructing a set is using a subquery as described in section 4.11.
4.8 Array Index Operator
It is possible to take a subsection or a single element from an array column, keyword or expression using the
index operator [index1,index2,...]. This syntax is similar to that used in Python or Glish. Similar to
Python a negative value can be given meaning counting from the end. However, a negative stride cannot be
given. Taking a single element can be done as:
  array[1, 2]
   array[-1, -1]                  last element
   array[1, some_expression]
Taking a subsection can be done as:
  array[start1:end1:incr1, start2:end2:incr2, ...]
If a start value is left out it defaults to the beginning of that axis. An end value defaults to the end of the
axis and an increment defaults to one. If an entire axis is left out, it defaults to the entire axis.
E.g., an array with shape [10,15,20] can be subsectioned as:
  [,,3]               resulting in an array of shape [10,15,1]
   [2:4, ::3, 2:15:2]  resulting in an array of shape [3,5,7]
                       (NB. shape is [2,5,7] for python style)
   [-1:-1,,]           last element of first axis, all elements other axes
The examples show that an index can be a simple constant (as it will usually be). It can also be an
expression which can be as complex as one likes. The expression has to result in a real value which will be
truncated to an integer. For ï¬xed shaped arrays checking if array bounds are exceeded is done at parse time. For variable shaped
arrays it can only be done per row. If array bounds are exceeded, an exception is thrown. In the future a
special undeï¬ned value will be assigned if bounds of variable shaped arrays are exceeded to prevent the
selection process from aborting due to the exception.
Note that the index operator will be applied directly to a column. This results in reading only the
required part of the array from the table column on disk. It is, however, also possible to apply it to a
subexpression (enclosed in parentheses) resulting in an array. E.g.
  arraycolumn[2,3,4] + 1
   (arraycolumn + 1)[2,3,4]
can both be used and have the same result. However, the ï¬rst form is faster, because only a
single element is read (resulting in a scalar) and 1 is added to it. The second form results in
reading the entire array. 1 is added to all elements and only then the requested element is taken.
From this example it should be clear that indexing an array expression has to be done with
care.
4.9 Units
Units can be given at many places in an expression; in fact, each subexpression can be ended with a unit
meaning that the subexpression result gets that unit or will be converted to that unit. A simple unit (only
letters) can always be given literally. A non-simple unit can be given literally if only containing digits,
underscores and/or dots (e.g., m2, fl_oz. or m.m). Otherwise the unit has to be enclosed in single or double
quotes (e.g., ’m/s’) or the backslash has to be used as escape character (e.g., \in or m\/s).
Arguments to functions like sin will be converted to the appropriate unit (radians) as needed. In a similar
way, the units of operands to operators like addition, will be converted as needed. An exception is thrown if a
unit conversion is not possible.
All units supported by module Quanta can be used. Note that the units are case sensitive. Most common
units use lowercase characters. A unit can be preceded by a scaling preï¬x (like k for kilo). Compound units
are created when multiplying or dividing values with units. The available units can be shown using the showunits command of the program taql.
Units can be given as follows:
If a column has a unit deï¬ned in column keyword QuantumUnits or UNIT, it automatically gets
that unit.
A constant can immediately be followed by a simple unit. e.g., 2deg.
The result of several expressions have an implicit unit. Constants given as positions are in radians (rad). Difference of 2 dates is in days (d). Inverse trigonometric functions like asin give radians.
When combining values with different units in e.g., an interval, a set, an addition, or a function like
min, the values are converted to the unit of the ï¬rst operand or argument with a unit.
Values without a unit have by default the unit of the ï¬rst operand or argument with a
unit.
  3mm-7cm          result is -67 mm
   3+3mm            result is 6 mm
   3mm<:<3cm        result is interval <3mm,30mm>
   [3,4cm,5]        result is [3cm, 4cm, 5cm]
   [5, 7cm, 8mm]    result is [5cm, 7cm, 0.8cm]
   [5, 7mm, 8cm]    result is [5mm, 7mm, 80mm]
   max(3mm,2cm)     result is 20 mm
   5 ’km/h’ + 1 ’m/s’      is 8.6 km/h
   iif(F,3min,30sec)       is 0.5 min
Similarly, operands of comparison operators and arguments of comparison functions (like near) are
converted to the unit of the ï¬rst operand or argument with a unit.
The result of a multiplication and division is a compound unit if both operands have a unit. Otherwise
it is the unit of the argument with a unit. Before TaQL supported units, it was needed to divide the TIME column in a MeasurementSet
by 86400 to convert it to days, so it could be compared with a given date/time. So, for
backward compatibility, a division of a value with unit s by a constant 86400 results in unit
d.
Division of units of an equal kind (e.g., km by m) results in a unitless value.
Division of a unitless value by a value with unit, results in the reciproke unit.
  1km/10m          result is 100
   1/2s             result is 0.5 ’(s)-1’
   1Hz + 1/2s       result is 1.5Hz
The result of functions like SUMSQR and SQRT is a compound unit if the argument has a unit. Note that
sqrt(2m) will fail, because the square root of a meter does not exist.
A (sub)expression can be followed by a simple or compound unit. If the subexpression
has no unit, it gets the given unit. Otherwise the resulting value is converted to the unit.
Note that some units can be the same as a reserved word (e.g., as or in). In that case it has to be
escaped or enclosed in quotes.
  COL \in          set/convert column COL to inch
   3 cm             result is 3 cm
   3 ’km/s’         result is 3 km/second
   3mm cm           result is 0.3 cm
   (3mm cm)m        result is 0.003 m
   (3+3) cm         result is 6 cm
   (3+3mm) cm       result is 0.6 cm
   [3,4,5]mm        result is [3mm, 4mm, 5mm]
   [3,4cm,5]mm      result is [30mm, 40mm, 50mm]
       Note: all values in the set first get the same unit cm
   asin(1)          result is pi/2 radians
   asin(1) deg      result is 90 degrees
   (3mm+7cm) m      result is 0.073 m
If a function argument is expected in a certain unit, values are converted as needed. For example,
arguments to functions sin and anycone are automatically converted to radians.
When adding or subtracting a value from a date, that value is converted to unit d (days).
Units will probably mostly be used in an expression in the WHERE clause or in a CALC command.
However, it is also possible to use a unit in the selection of a column in the SELECT clause. For
example:
  select TIME d as TIMED from my.ms
In such a case the selection is an expression and the unit is stored in the column keywords. Thus in this
example, TIME is stored in a column TIMED with keyword QuantumUnits=d and the values are converted to
days.
4.10 Functions
More than 100 functions exist to operate on scalar and/or array values. Some functions have two names. One
name is the CASA/Glish name, while the other is the name as used in SQL. In the following tables the
function names are shown in uppercase, while the result and argument types are shown in lowercase. Note,
however, that function names are case-insensitive. Furthermore it is possible to have user deï¬ned functions that are dynamically loaded from a shared library.
In section Writing user deï¬ned functions it is explained how to write user deï¬ned functions.
A set of standard UDFs exists dealing with Measure conversions, for example to convert J2000 to
apparent. Another set of UDFs deals with values and relations in MeasurementSets and Calibration
Tables.
Sets, and in particular subqueries, can result in a 1-dim array. This means that the functions accepting an
array argument can also be used on a set or the result of a subquery.
4.10.1 String functions
These functions can be used on a scalar or an array argument.
integer STRLENGTH(string), integer LEN(string) Returns the number of characters in a string (trailing whitespace is signiï¬cant).
string UPCASE(string), string UPPER(string) Convert to uppercase.
string DOWNCASE(string), string LOWER(string) Convert to lowercase.
string CAPITALIZE(string) Capitalize a string (make ï¬rst letter uppercase).
string LTRIM(string) Removes leading whitespace.
string RTRIM(string) Removes trailing whitespace.
string TRIM(string) Removes leading and trailing whitespace.
string SUBSTR(string, integer ST, integer N) Returns a substring starting at the 0-based position ST with a length of at most N characters. If
the string argument is an array of strings, an array with the substring of each string is returned.
The arguments ST and N have to be scalar values. If ST is negative, it counts from the end (a
la Python). N and the resulting ST will be set to 0 if negative.
string REPLACE(string SRC, PATTERN, string REPL) Replaces all occurrences of PATTERN in SRC by REPL and returns the result. REPL can
be omitted and defaults to the empty string. If the ï¬rst argument is an array of strings, each
element in the array is replaced. The arguments PATTERN and REPL have to be scalar values.
PATTERN can be a string or a regular expression (see below). For example: REPLACE("abcdab", "ab") results in cd REPLACE("abcdab", REGEX("^ab"), "xyz") results in xyzcdab
4.10.2 Regex functions
Apart from using regex/pattern constants, it is possible to use functions to form a regex or pattern. These
functions can only be used on a scalar argument.
regex REGEX(string) Handle the given string as a regular expression.
regex PATTERN(string) Handle the given string as a UNIX ï¬lename-like pattern and convert it to a regular expression.
regex SQLPATTERN(string) Handle the given string as an SQL-style pattern and convert it to a regular expression.
A regex formed this way can only be used in a comparison == or !=. E.g. object == pattern(’3C*’) to ï¬nd all 3C objects in a catalogue.
A few remarks:
The regex/pattern functions and operator LIKE work on any string, thus they can be used with
any string expression.
A Regex is case sensitive. One should use function upcase or downcase on the string to test to
make it case insensitive or use the i qualiï¬er on a regex constant.
Usually a regex/pattern must match the full string, thus not part of it. However, one can use
the m// regex constant to do partial matching. Thus something like m/xx/ matches all strings
containing xx. Of course, regex(’.*xx.*’) can also be used. In this way the m// regex works
the same as in languages like Perl, Python, and Glish.
4.10.3 Date/time functions
These functions make it possible to handle dates/times and can be used on a scalar or an array argument.
The syntax of a date/time string or constant is explained in section 4.3.7.
DateTime DATETIME(string) Parse the string and convert it to a DateTime value.
DateTime MJDTODATE(real) The real value, which has to be a MJD (Modiï¬edJulianDate), is converted to a DateTime.
double MJD(DateTime) Get the DateTime as MJD (Modiï¬edJulianDate) in days.
DateTime DATE(DateTime) Get the date (i.e., remove the time part). This function is needed in something like: DATE(column) == 12Feb1997 if the column contains date/times with times0.
double TIME(DateTime) Get the time part of the day. It is converted to radians to be compatible with the internal
representation of times/positions. In that way the function can easily be used as in: TIME(date) 12h
integer YEAR(DateTime) Get the year (which includes the century).
integer MONTH(DateTime) Get the month number (1-12).
integer DAY(DateTime) Get the day number (1-31).
integer WEEK(DateTime) Get the week number in the year (0 ... 53). Note that week 1 is the week containing Jan 4th.
integer WEEKDAY(DateTime), integer DOW(DateTime) Get the weekday number (1=Monday, ..., 7=Sunday).
string CDATETIME(DateTime), string CTOD(DateTime) Get the DateTime as a string like YYYY/MM/DD/HH:MM:SS.SSS.
string CDATE(DateTime) Get the date part of a DateTime as a string like DD-MMM-YYYY.
string CTIME(DateTime) Get the time part of a DateTime as a string like HH:MM:SS.SSS.
string CMONTH(DateTime) Get the abbreviated name of the month (Jan ... Dec).
string CWEEKDAY(DateTime), string CDOW(DateTime) Get the abbreviated name of the weekday (Mon ... Sun).
All functions can be used without an argument in which case the current date/time is used. e.g., DATE() results
in the current date. It is possible to give a string argument instead of a date. In this case the string is parsed and converted to a
date (i.e., the function DATETIME is used implicitly). Note that the function STR discussed in the next section can also be used for pretty-printing a date/time. It
gives more control over the number of decimals and date format.
4.10.4 Pretty printing functions
Angles (scalar or array) can be returned as strings in HMS and/or DMS format. Currently, they are always
formatted with 3 decimals in the seconds.
string HMS(real) Return angle(s) like 12h34m56.789
string DMS(real) Return angle(s) like 12d34m56.789
string HDMS(realarray) Return angles like 12h34m56.789 (even elements) and 12d34m56.789 (odd elements). It is useful
for arrays containing RA,DEC values.
The functions mentioned above and the date/time functions in the previous subsection can format a value in a
predeï¬ned way only. The STRING (shorthand STR) function makes it possible to convert values to strings using an optional format
string or width.precision value. It also makes it possible to format dates, times, and angles in a variety of
ways.
The value can be of any type (except Regex) and can be a scalar or array. The optional format must be a
scalar string or numeric value. If no format is given, an appropriate default format will be used.
A numeric format value deï¬nes the width and/or precision. For example:
  8        defines width 8 and default precision
   20.12    defines width 20 and precision 12
   .8       defines precision 8 and default width
In this way precision represents all digits, not only the ones behind the decimal point. A default width
or precision is used if not given.
A string format value can contain a printf-style format string, which must include the %-sign. Note
that the real and imaginary part of a complex value are formatted separately, so such a format string
needs to contain a format speciï¬er for both parts. See printf reference for possible format speciï¬ers. For
example:
  %10d     decimal with width 10
   %010d    decimal with width 10 and filled with zeroes
   %f+%fi   to format a complex value as a+bi
Apart from a printf-style format string, it is also possible to deï¬ne a string to format
date/time and angle values (which are automatically converted to radians if containing units).
Such a format string contains one or more format values as deï¬ned in class MVTime. A vertical bar
(with optional whitespace) must be used as separator. A string part can be a numeric value deï¬ning the
precision of the time/angle. Default precision is 6 (thus hh:mm:ss). The optional time/angle formats
and modiï¬ers are:
Format
Description
YMD
yyyy/mm/dd/hh:mm:ss.sss
YMD_ONLY
YMD without the time (same as YMDNO_TIME)
DMY
dd-Mon-yyyy/hh:mm:ss.sss
FITS
yyyy-mm-ddThh:mm:ss.sss
BOOST
the same as DMYUSE_SPACE
NO_H, NO_D
suppress the output of hours (or degrees): useful for offsets
NO_HM, NO_DM
suppress the degrees and minutes
CLEAN
suppress leading or trailing periods or colons if not all time/angle parts
are printed (e.g., when giving NO_H or 4 decimals)
DAY
precede the output with Day- (e.g., Wed-)
NO_TIME
suppress printing of time
ANGLE
+ddd.mm.ss.ttt
TIME
hh:mm:ss.ttt
USE_SPACE
use a space between date and time (and day and date)
DIG2
get angle/time in range -90:+90 or -12:+12
LOCAL
local time; in FITS mode append time zone as +hh:mm
For example:
  YMD             format as YYYY/MM/DD/HH:MM:SS
   DMY|NO_TIME     format as DD-MMM-YYYY
   DMY | DAY | 8   format as Thu-DD-MMM-YYY/HH:MM:SS.SS
   TIME            format a datetime or angle as HH:MM:SS
   ANGLE|9         format an angle as DD.MM.SS.SSS
If such a format string contains an invalid part, it is assumed that the entire string is a printf-style
format string.
4.10.5 Comparison functions
The exact comparison of floating point values is quite tricky. Two functions make it possible to compare 2
double or complex values with a tolerance. They can be used on scalar and array arguments (and a mix of
them). The tolerance must be a scalar though. Note that operator  = is the same as NEAR with a tolerance of 1e-5.
bool NEAR(numeric val1, numeric val2, double tol) Tests in a relative way if a value is near another. Relative means that the magnitude of the
numbers is taken into account. It returns abs(val2 - val1)/max(abs(val1),abs(val2)) < tol. If tol<=0, it returns val1==val2. If either val is 0.0, it takes care of area around the minimum
number that can be represented. The default tolerance is 1.0e-13.
bool NEARABS(numeric val1, numeric val2, double tol) Tests in an absolute way if a value is near another. Absolute means that the magnitude of the
numbers is not taken into account. It returns abs(val2 - val1) < tol. The default tolerance is 1.0e-13.
bool ISNAN(numeric val) Tests if a numeric value is a NaN (not-a-number).
bool ISINF(numeric val) Tests if a numeric value is inï¬nite (positive or negative).
bool ISFINITE(numeric val) Tests if a numeric value is a ï¬nite number (not NaN or inï¬nite).
4.10.6 Mathematical functions
Standard mathematical can be used on scalar and array arguments (and a mix of them).
double PI() Return the value of pi.
double E() Return the value of e (is equal to EXP(1)).
double C() Return the value of the speed of light (with unit m/s).
dnumeric SIN(numeric)
dnumeric SINH(numeric)
double ASIN(real)
dnumeric COS(numeric)
dnumeric COSH(numeric)
double ACOS(real)
double TAN(real)
double TANH(real)
double ATAN(real)
double ATAN2(real y, real x) Return ATAN(y/x) in correct quadrant.
dnumeric EXP(numeric)
dnumeric LOG(numeric) Natural logarithm.
dnumeric LOG10(numeric)
dnumeric POW(numeric, numeric) The same as operator **.
numeric SQUARE(numeric), numeric SQR(numeric) The same as **2, but much faster.
dnumeric SQRT(numeric)
complex COMPLEX(real, real)
dnumeric CONJ(numeric)
double REAL(numeric) Real part of a complex number. Returns argument if real.
double IMAG(numeric) Imaginary part of a complex number. Returns 0 if argument is real.
real NORM(numeric)
real ABS(numeric), real AMPLITUDE(numeric)
double ARG(numeric), double PHASE(numeric)
numeric MIN(numeric, numeric)
numeric MAX(numeric, numeric)
real SIGN(real) Return -1 for a negative value, 0 for zero, 1 for a positive value.
real ROUND(real) Return the rounded value of the number. Negative numbers are rounded in an absolute way.
e.g., ROUND(-1.6) = -2.
real FLOOR(real) Works towards negative inï¬nity. e.g., FLOOR(-1.2) = -2.
real CEIL(real) Works towards positive inï¬nity.
real FMOD(real, real) The same as operator %.
Note that the arguments or results of the trigonometric functions are in radians. They are converted
automatically if units are given.
4.10.7 Array to scalar reduce functions
The following functions reduce an array to a scalar. They are meant for an array, but can also be used for a
scalar.
bool ANY(bool) Is any element true?
bool ALL(bool) Are all elements true?
integer NTRUE(bool) Return number of true elements.
integer NFALSE(bool) Return number of false elements.
numeric SUM(numeric) Return sum of all elements.
numeric SUMSQUARE(numeric), numeric SUMSQR(numeric) Return sum of all squared elements.
numeric PRODUCT(numeric) Return product of all elements.
real MIN(real) Return minimum of all elements.
real MAX(real) Return maximum of all elements.
dnumeric MEAN(numeric), dnumeric AVG(numeric) Return mean of all elements.
double VARIANCE(real) Return variance (the sum of (a(i) - mean(a))**2/(nelements(a) - 1).
double STDDEV(real) Return standard deviation (the square root of the variance).
double AVDEV(real) Return average deviation. (the sum of abs(a[i] - mean(a))/nelements(a).
double RMS(real) Return root-mean-squares. (the square root of the sum of (a(i)**2)/nelements(a).
double MEDIAN(real) Return median (the middle element). If the array has an even number of elements, the mean of
the two middle elements is returned.
double FRACTILE(real, doublescalar fraction) Return the value of the element at the given fraction. Fraction 0.5 is the same as the median,
but no mean of the two middle elements is taken.
4.10.8 Array to array reduce functions
These functions reduce an array to a smaller array by collapsing the given axes using the given function. The
axes are the last argument(s). They can be given in two ways: - As a single set argument; for example, maxs(ARRAY,[1,2]) - As individual scalar arguments; for example, maxs(ARRAY,1,2) For example, using MINS(array,0,1) for a 3-dim array results in a 1-dim array where each value is the
minimum of each plane in the cube. It is important to note that the interpretation of the axes numbers depends on the style being
used. e.g., when using glish style, axes numbers are 1-based and in Fortran order, thus axis 1 is
the most rapidly varying axis. When using python style, axis 0 is the most slowly varying axis.
Axes numbers exceeding the dimensionality of the array are ignored. For example, maxs(ARRAY,[1:10])
works for arrays of virtually any dimensionality and results in a 1-dim array. The function names are the ’plural’ forms of the functions in the previous section. They can only be used for
arrays, thus not for scalars.
bool ANYS(bool) Is any element true?
bool ALLS(bool) Are all elements true?
integer NTRUES(bool) Return number of true elements.
integer NFALSES(bool) Return number of false elements.
numeric SUMS(numeric) Return sum of elements.
numeric SUMSQUARES(numeric), numeric SUMSQRS(numeric) Return sum of squared elements.
numeric PRODUCTS(numeric) Return product of elements.
real MINS(real) Return minimum of elements.
real MAXS(real) Return maximum of elements.
dnumeric MEANS(numeric), dnumeric AVGS(numeric) Return mean of elements.
double VARIANCES(real) Return variance (the sum of (a(i) - mean(a))**2/(nelements(a) - 1).
double STDDEVS(real) Return standard deviation (the square root of the variance).
double AVDEVS(real) Return average deviation. (the sum of abs(a(i) - mean(a))/nelements(a).
double RMSS(real) Return root-mean-squares. (the square root of the sum of (a(i)**2)/nelements(a).
double MEDIANS(real) Return median (the middle element). If the array has an even number of elements, the mean of
the two middle elements is returned.
double FRACTILES(real, doublescalar fraction) Return the value of the element at the given fraction. Fraction 0.5 is the same as the median.
4.10.9 Array downsampling functions
These functions are a generalization of the functions in the previous section. They downsample an array by
taking, say, the mean of every n*m elements. The functions in the previous section downsample by taking the
mean of a full line or plane, etc. The most useful one is probably calculating the boxed mean, but the other
ones can be used similarly. The width of each window axis has to be given. Missing axes default to 1.
Similarly to the partial reduce functions described above, the axes can be given as scalars or as a set.
For example, BOXEDMEAN(array,3,3) calculates the mean in each 3x3 box. At the end of an axis the box
used will be smaller if it does not ï¬t integrally. The functions can only be used for arrays, thus not for scalars.
bool BOXEDANY(bool) Is any element true?
bool BOXEDALL(bool) Are all elements true?
double BOXEDMIN(real) Return minimum of elements.
double BOXEDMAX(real) Return maximum of elements.
dnumeric BOXEDMEAN(numeric), dnumeric BOXEDAVG(numeric) Return mean of elements.
double BOXEDVARIANCE(real) Return variance (the sum of (a(i) - mean(a))**2/(nelements(a) - 1).
double BOXEDSTDDEV(real) Return standard deviation (the square root of the variance).
double BOXEDAVDEV(real) Return average deviation. (the sum of abs(a(i) - mean(a))/nelements(a).
double BOXEDRMS(real) Return root-mean-squares. (the square root of the sum of (a(i)**2)/nelements(a).
double BOXEDMEDIAN(real) Return median (the middle element).
4.10.10 Array functions operating in running windows
These functions transform an array into an array with the same shape by operating on a rectangular
window around each array element. The most useful one is probably calculating the running
median, but the other ones can be used similarly. The half-width of each window axis has to be
given; the full width is 2*halfwidth + 1. Missing axes default to a half-width of 0. Similarly
to the partial reduce functions described above, the axes can be given as scalars or as a set.
For example, RUNNINGMEDIAN(array,1,1) calculates the median in a 3x3 box around each array element. See
the examples how it is applied to an image. In the result the edge elements (i.e., the elements where no full window can be applied) are set to 0 (or
False). The functions can only be used for arrays, thus not for scalars.
bool RUNNINGANY(bool) Is any element true?
bool RUNNINGALL(bool) Are all elements true?
double RUNNINGMIN(real) Return minimum of elements.
double RUNNINGMAX(real) Return maximum of elements.
dnumeric RUNNINGMEAN(numeric), dnumeric RUNNINGAVG(numeric) Return mean of elements.
double RUNNINGVARIANCE(real) Return variance (the sum of (a(i) - mean(a))**2/(nelements(a) - 1).
double RUNNINGSTDDEV(real) Return standard deviation (the square root of the variance).
double RUNNINGAVDEV(real) Return average deviation. (the sum of abs(a(i) - mean(a))/nelements(a).
double RUNNINGRMS(real) Return root-mean-squares. (the square root of the sum of (a(i)**2)/nelements(a).
double RUNNINGMEDIAN(real) Return median (the middle element).
4.10.11 Type conversion functions
Explicit type conversions can be done using one of the functions below. They can operate on scalars and
arrays.
integer INT(numeric or bool or string) Convert the argument to an integer. A real number is truncated (-10.9 results in -10). For a
complex number the truncated real part is taken. A bool is converted to 0 (False) or 1 (True).
It does not check if a string represents a valid integer. It is interpreted until the ï¬rst non-valid
character, so a string containing a floating point value is truncated.
double REAL(numeric or bool or string) Convert the argument to a real number. For a complex number the real part is taken. A bool is
converted to 0 (False) or 1 (True). It does not check if a string represents a valid floating point
value. A string is interpreted until the ï¬rst non-valid character.
complex COMPLEX(real,real) Form a complex number from the given real and imaginary part.
complex COMPLEX(string) Convert the string to a complex number. The number can be given like (1,2) or 1+2i. In fact,
any separator (except whitespace) between real and imaginary part is possible. It does not check
if a string represents a valid complex value. The string is interpreted until the ï¬rst non-valid
character, so the last character can be any character (e.g., also j).
bool BOOL(anytype) Convert the value to a bool. A numeric type (or date) results in False if the value is 0, otherwise
True. A string is case-insensitive. False, F, No, N, -, or 0 results in False, otherwise True.
4.10.12 Array creation functions
anytypearray ARRAY(anytype,shape) This function creates an unmasked array of the given type and shape. The shape is given in the
last argument(s). It can be given in two ways: - As a single set argument; for example, array(0,[3,4]) - As individual scalar arguments; for example, array(0,3,4) The ï¬rst argument gives the values the array is ï¬lled with. It can be a scalar or an array of any
shape. To initialize the created array, the value array is flattened to a 1D array. Its successive
values are stored in the created array. If the new array has more values than the value array, the
value array is reset to its beginning and the process continues. Note that a masked array can be created from an (unmasked) array and a mask using the
brackets operator like ARRAY[MASK].
anytypearray MARRAY(anytypearray,boolarray) This function offers another way to create a masked array. The mask must be given in the second
argument; its shape must be the same as the shape of the data array.
anytypearray NULLARRAY(anytype) This function creates a null array. Its data type is determined by the data type of the argument.
The argument value itself is not used. It is mainly meant for test purposes.
anytypearray RESIZE(anytypearray,shape[,mode]) This function resizes an array to the given shape and copies the values. The optional mode
argument determines how the values are copied. If the argument is not given, the new shape is
arbitrary and the dimensionality can change. The values are copied to the same index in the
new array. If an axis gets larger, the new values are set to 0 (or an empty string). If mode is given, the new shape is much more restricted. The dimensionality cannot change
and each axis has to be a multiple of the old one. If mode=0, copying the values is done in an
upsampling way. E.g., if a new axis is twice the length of the old one, values 1,2,3 are copied as
1,1,2,2,3,3. A good use case is applying the flags of averaged data to the original data. If mode=1,
the values in the example above are copied repeatedly as 1,2,3,1,2,3. By giving the mode as a
set, it is possible to specify the mode per axis, but that is quite esoteric.
anytypearray TRANSPOSE(anytypearray[,axes]) This function transposes an N-dim array. If no axes are given, the array is fully transposed (thus
all axes are reversed). Axes can be speciï¬ed meaning that those axes will become the ï¬rst axes
in the output array. Non-given axes follow thereafter in their natural order. A possible mask is transposed as well.
anytypearray DIAGONAL(anytypearray[,firstaxis[,diag]]) This function takes the diagonal of 2-dim subarrays in an N-dim array resulting in an array
with 1 dimension less. For a 2-dim array, it is simply the diagonal of the matrix. For a higher
dimensional array, it takes the diagonal of each matrix deï¬ned by firstaxis and firstaxis+1.
e.g., in a 3-dim array the diagonals of each XY-plane can be taken. The default for ï¬rstaxis is
0. The diag argument tells which diagonal has to be taken. The default 0 means the main diagonal.
A negative value means below the main diagonal, while positive means above the main diagonal.
anytypearray ARRAYDATA(anytype) This function returns the array without a mask, thus removes the mask. If the operand is a
scalar, it returns a 1-dim array with one element.
anytypearray NEGATEMASK(anytype) This function returns the array with the negated mask. If the array has no mask, it returns the
array with a mask of all Trues. If the operand is a scalar, it returns a 1-dim array with one
element.
anytypearray REPLACEMASKED(anytype, anytype) This function replaces the masked elements in the ï¬rst argument by the corresponding value
in the second argument (which can be a scalar value). If the ï¬rst argument has no mask, the
function is a no-op.
anytypearray REPLACEUNMASKED(anytype, anytype) This function replaces the unmasked elements in the ï¬rst argument by the corresponding value
in the second argument (which can be a scalar value). If the ï¬rst argument has no mask, the
function is a no-op.
boolarray ARRAYMASK(anytype), boolarray MASK(anytype) This function returns the mask of an array. If the array has no mask, it returns a boolean array
of the same shape with all values set to False. If the operand is a scalar, it returns an 1-dim
array with one False element.
anytypearray FLATTEN(anytype) This function flattens an N-dim array to a 1-dim array keeping the unmasked elements only. If
the operand is a scalar, it returns a 1-dim array with one element.
4.10.13 Aggregate functions
The GXXX aggregate functions calculate an aggregated value for all rows in a group, usually deï¬ned with a
GROUPBY clause. For example, when grouping in TIME, an aggregate function like GNTRUE(FLAG) counts
per time slot the number of flagged data points. Aggregate functions can only be used in the SELECT and
the HAVING clause. Most functions listed below reduce the values in a group to a scalar value, also if the value in a
row is an array (as in the GNTRUE example above). The arrays in a group can have different
shapes.
However, there are several aggregate functions returning an array as done by the last three
functions (GHIST, GAGGR, and GROWID) shown below. Furthermore, most scalar functions
have a plural form (e.g., GNTRUES) returning an array. They are described at the end of this
section.
Note that the aggregate function names differ from their SQL counterparts; they all have the preï¬x G,
because TaQL functions like MAX already exist for array operations. This naming scheme also makes it more
clear which TaQL functions are aggregate functions.
A technical detail is how aggregate functions are implemented. TaQL walks sequentially through a
table. Non-lazy functions operate directly on the value in a row making the table access purely
sequential. It requires that the results of all groups are held in memory. For some functions, in
particular GAGGR, this could lead to a very high memory usage. Therefore, some functions are
implemented in a lazy way. They keep the row numbers of a group and access the data when the
aggregated result of a group is needed. In this way only the data of a single group needs to be held
in memory, but the access to the table might be non-sequential making it somewhat slower.
Currently, only GAGGR and the User Deï¬ned aggregate functions are implemented in a lazy
way.
integer GCOUNT(), integer GCOUNT(*) Return the number of rows per group.
integer GCOUNT(columnname) Return the number of rows per group for which the column has a value. Note that only a column
containing variable sized arrays can contain empty cells.
anytype GFIRST(anytype) Return the ï¬rst value of an expression in the group. The values of a column not mentioned in
the GROUPBY clause, might differ. This function can be used to return the value of the ï¬rst
row in the group.
anytype GLAST(anytype) Return the last value of the group (is similar to GFIRST). Note this function is implicitly used if an expression without aggregate function is used in a
group.
bool GANY(bool) Is any element true?
bool GALL(bool) Are all elements true?
integer GNTRUE(bool) Return number of true elements.
integer GNFALSE(bool) Return number of false elements.
numeric GSUM(numeric) Return sum of all elements.
numeric GSUMSQUARE(numeric), numeric GSUMSQR(numeric) Return sum of all squared elements.
numeric GPRODUCT(numeric) Return product of all elements.
real GMIN(real) Return minimum of all elements.
real GMAX(real) Return maximum of all elements.
dnumeric GMEAN(numeric), dnumeric GAVG(numeric) Return mean of all elements.
double GVARIANCE(real) Return variance (the sum of (a(i) - mean(a))**2/(nelements(a) - 1).
double GSTDDEV(real) Return standard deviation (the square root of the variance).
double GAVDEV(real) Return average deviation. (the sum of abs(a[i] - mean(a))/nelements(a).
double GRMS(real) Return root-mean-squares. (the square root of the sum of (a(i)**2)/nelements(a).
double GMEDIAN(real) Return median (the middle element). If the array has an even number of elements, the mean of
the two middle elements is returned.
double GFRACTILE(real, doublescalar fraction) Return the value of the element at the given fraction. Fraction 0.5 is the same as the median.
double GHIST(real, intscalar nbin, realscalar start, realscalar end) Return the histogram of the data using the given number of bins. The histogram contains an
extra bin at the beginning and the end for the outliers. If the rows in the group contain arrays,
they can have variable shapes.
anytypearray GAGGR(anytype), anytypearray GSTACK(anytype) Stack the row values in a group to form an array where the row is the slowest varying axis
(similar to numpy’s dstack). Thus if the column contains scalar values, the result is a vector.
Otherwise it is an array whose dimensionality is one higher. It requires that all arrays in a group
have the same shape. Note that this function can be very useful for arrays, because it makes it possible to use partial
reduce functions like medians to calculate the medians along arbitrary axes.
integerarray GROWID() Return the row numbers of the rows in the group.
Most functions above have a plural counterpart. They calculate the aggregated value per array index, thus the
result has the same shape as the arrays in the group. Similar to function GAGGR, they require that all
arrays in a group have the same shape. For instance, for a MeasurementSet the expression GMEANS(DATA) calculates the mean in a group per
channel/polarization. Not only it is a shorthand for MEANS(GAGGR(DATA), 0), but it usually works faster
because, unlike GAGGR, it is non-lazy. The functions available are:
  GANYS   GALLS      GNTRUES     GNFALSES
   GMINS   GMAXS
   GSUMS   GPRODUCTS  GSUMSQRS    GSUMSQUARES
   GMEANS  GAVGS      GVARIANCES  GSTDDEVS  GRMSS
4.10.14 Miscellaneous functions
bool ISNULL(anytype) Return True if the argument value is a null array.
bool ISDEFINED(anytype) Return False if the array value in the current row is undeï¬ned (is null). It makes it possible to
test if a cell in a column with variable shaped arrays contains an array. Furthermore, it can be
used to test if a ï¬eld in a record is deï¬ned. Note that function ISNULL can also be used to test for an undeï¬ned array in a row.
bool sh.ISCOLUMN(string) Return False if no column with the given name exists in the table with the shorthand given
before the function name. If no table is given, the ï¬rst one will be used.
bool sh.ISKEYWORD(string) Return False if no keyword with the given name exists in the table with the shorthand given
before the function name. If no table is given, the ï¬rst one will be used. The keyword name can
be given as described in section 4.5, thus the name of a table keyword or column keyword or a
nested ï¬eld can be speciï¬ed.
integer NELEMENTS(anytype), integer COUNT(anytype) Return number of elements in an array (1 for a scalar).
integer NDIM(anytype) Return dimensionality of an array (0 for a scalar).
integerarray SHAPE(anytype) Return shape of an array (returns an empty array for a scalar).
integer ROWNUMBER(), integer ROWNR() Return the row number being tested (ï¬rst row is row number 0 or 1 depending on the style
used). In combination with function RAND it can, for instance, be used to select arbitrary rows from
a table.
integer ROWID() Return the row number in the original table. This is especially useful for returning the result of a
selection of a subtable of a Casacore measurement set (see also subqueries in 4.11 and examples
in section 12.1).
double RAND() Return (per table row) a uniformly distributed random number between 0 and 1 using a
Multiplicative Linear Congruential Generator. The seeds for the generator are deduced from the
current date and time, so the results are different from run to run. The function can, for instance, be used to select a random subset from a table.
double ANGDIST(arg1,arg2), double ANGULARDISTANCE(arg1,arg2) Return the angular distance (in radians) between the positions in arg1 and arg2. Both arguments
have to be numeric arrays containing an even number of values. Two subsequent values give
the RA and DEC (or longitude and latitude) of positions on a sphere. The result is a 1-dim
array containing the angular distance between corresponding positions in arg1 and arg2. If
either array contains only one position, the result is the distance between that position and each
position in the other array. If both arguments contain only 2 values, the result is a scalar. For
example: angdist(PHASE_DIR[0,], [12h13m45,4d21m39.4, 12h13m49,10d8m4]) returns an array with shape [2] containing the angular distance between the phase center of the
ï¬eld and the two positions given.
double ANGDISTX(arg1,arg2), double ANGULARDISTANCEX(arg1,arg2) Same as above, but the result is a 2-dim array giving the distance between each position in
the ï¬rst argument and each position in the second argument. Only if both arguments contain a
single position, the result is a scalar.
anytype IIF(cond,arg1,arg2) This is a special funtion which operates like the ternary ?: operator in C++. If all arguments
are scalars, the result is a scalar, otherwise an array. In the latter case possible scalar arguments
are virtually expanded to arrays. IIF evaluates the condition for each element. If True, it takes
the corresponding element of arg1, otherwise of arg2. If one of the input arrays has a mask, the output array will also have a mask. Each output mask
element value is the logical OR of the condition mask element value and the mask value of the
element taken from arg1 or arg2.
4.10.15 Cone search functions
Cone search functions make it possible to test if a source is within a given distance of a given sky position.
The expression
  cos(0d1m) < sin(52deg) * sin(DEC) +
               cos(52deg) * cos(DEC) * cos(3h30m - RA)
could be used to test if sources with their sky position deï¬ned in columns RA and DEC are within 1 arcmin of
the given sky position. The cone search functions implement this expression making life much easier for the user. Because they can
also operate on arrays of positions, searching in multiple cones can be done simultaneously. That makes it
possible to ï¬nd matching source positions in two catalogues as shown in an example at the end of this
section.
The arguments of all functions are described below. All of them have to be given in radians. However,
usually one does not need to bother because TaQL makes it possible to specify positions in many formats
automatically converted to radians.
SOURCES is a set or array giving the positions of one or more sources (e.g., in equatorial coordinates) to
be tested. Normally these are columns in a table. Where argument name SOURCE is mentioned
below, only a single source can be used, otherwise multiple sources. For example: [RA,DEC] for scalar columns RA and DEC. SKYPOS for a column SKYPOS containing 2-element vectors with RA and DEC.
CONES is a set or array giving the center positions and radii of one or more cones (e.g., as
RA,DEC,radius). Usually the user will specify it as constants. For example: [12h13m54, -5.3.34, 0d1m] for a single cone. [12h13m54, -5.3.34, 0d1m, 1h2m3, 4.5.6, 0d1m] for two cones.
CONEPOS is a set or array giving the positions of one or more cone centers (e.g., as RA,DEC).
RADII is a scalar, set or array giving one or more radii. Each radius is applied to all positions in
CONEPOS. Specifying a cone as CONEPOS,RADIUS is easier than specifying it as CONES if the same
radius has to be used for multiple cones. For example: [12h13m54, -5.3.34, 1h2m3, 4.5.6], 0d1m is the same as the second CONES example above.
The following cone search functions are available.
bool ANYCONE(SOURCE,CONES) Return T if the source is contained in at least one of the cones. Operator INCONE is a synonym.
So ANYCONE(SOURCE,CONES) is the same as SOURCE INCONE CONES.
bool ANYCONE(SOURCE,CONEPOS,RADII) It does the same as above.
integer FINDCONE(SOURCES,CONES) Return the index of the ï¬rst cone containing the source. If a single source is given, the result is
a scalar. If multiple sources are given, the result is an array with the same shape as the source
array.
integer FINDCONE(SOURCES,CONEPOS,RADII) It does the same as above. Note that in this case each radius is applied to each cone, so the
resulting index array is a combination of the two input arrays (with the radius as the most
rapidly varying axis).
bool CONES(SOURCES,CONES) Return a 2-dim bool array. The length of the most rapidly varying axis is the number of cones.
The length of the other axis is the number of sources. When using python style, element (i,j)
in the resulting array is T if source i is contained in cone j.
bool CONES(SOURCES,CONEPOS,RADII) It does the same as above. However, the result is a 3-dim array with the radii as the most rapidly
varying axis, cones as the next axis, and sources as the slowest axis.
Please note that ANYCONE(SOURCE,CONES) does the same as any(CONES(SOURCE,CONES)), but is faster because it
stops as soon as a cone is found. Function CONES makes it possible to do catalogue matching. For example, to ï¬nd sources matching other
sources in the same catalogue (within a radius of 10 arcseconds):
  CALC CONES([RA,DEC],
              [SELECT FROM table.cat GIVING [RA,DEC]], 0d0m10)
      FROM table.cat
Note that in this example the SELECT clause returns an array with positions which are used as the cone
centers. So each source in the catalogue is tested against every source. It makes it an N-square operation,
thus potentially very expensive. The result is a 4-dim boolean array with shape (in glish style)
[1,nrow,1,nrow] which can be processed in Glish. Please note that the CONES function results for each row
in a array with shape [1,nrow,1]. The query can be done with multiple radii, for example also with 1 arcsecond and 1 arcminute.
  CALC CONES([RA,DEC],
        [SELECT FROM table.cat GIVING [RA,DEC]], [0d0m1, 0d0m10, 0d1m])
      FROM table.cat
resulting in an array with glish shape [3,nrow,1,nrow]. In this way one can get a better indication how
close sources are to the cone centers.
4.10.16 User deï¬ned functions
TaQL can be extended with so-called User Deï¬ned Functions (UDF). These are dynamically loaded
functions, either written in C++ or in Python. In TaQL the name of a UDF written in C++ consists of the
name of the library (without lib preï¬x and extension) followed by a dot and the function name. For
example:
  meas.hadec(...)
denotes function hadec in shared library libmeas.so or libcasa_meas.so. For OS-X the extension .dylib
will be used. The physical shared library name must be fully lowercase, but the UDF name used in TaQL is
case-insensitive. The name of a UDF written in Python is like py.module.func where the module part
is optional. In the USING STYLE clause it is possible to deï¬ne synonyms for the UDF library
names. By default, mscal is deï¬ned as a synonym for derivedmscal and py as a synonym for
pytaql.
Usually a UDF will operate on the arguments given to the function and will not itself operate on a table
given in a query command. However, some UDFs (most notably the mscal ones) do not have arguments, but
operate directly in a speciï¬c way on a table. Normally they use the ï¬rst table given in the FROM
clause, but the UDF name can be preceded by a table shorthand to specify another table. For
example:
  select t1.mscal.ha1(), t2.mscal.ha1() from my1.ms t1, my2.ms t2
to get the hourangle from two different tables. Of course, both tables need to have the same number of rows.
Note that UDFs not directly operating on a table, will ignore a shorthand.
The Casacore package comes with several predeï¬ned UDFs in library libcasa_derivedmscal. It contains
four groups of UDFs, all operating on a MeasurementSet and several on a CASA calibration table (both old
and new format). Although the library is called derivedmscal, for ease of use it is possible to use the synonym
mscal.
Get derived values The ï¬rst group calculates derived values like hourangle and azimuth for the ï¬rst or second antenna of a
baseline. For calibration tables, where a row contains a single antenna, functions like PA1 are the same as
PA2. All angles are returned in radians.
double MSCAL.HA() gives the hourangle of the array center (observatory position).
double MSCAL.HA1() gives the hourangle of ANTENNA1.
double MSCAL.HA2() gives the hourangle of ANTENNA2.
double MSCAL.HADEC() gives the topocentric hourangle/declination of the array center (observatory position).
double MSCAL.HADEC1() gives the topocentric hourangle/declination of ANTENNA1.
double MSCAL.HADEC2() gives the topocentric hourangle/declination of ANTENNA2.
doublearray MSCAL.AZEL() gives the topocentric azimuth/elevation of the array center (observatory position).
doublearray MSCAL.AZEL1() gives the topocentric azimuth/elevation of ANTENNA1.
doublearray MSCAL.AZEL2() gives the topocentric azimuth/elevation of ANTENNA2.
double MSCAL.LAST() gives the local sidereal time of the array center.
double MSCAL.LAST1() gives the local sidereal time of ANTENNA1.
double MSCAL.LAST2() gives the local sidereal time of ANTENNA2.
double MSCAL.PA1() gives the parallactic angle of ANTENNA1.
double MSCAL.PA2() gives the parallactic angle of ANTENNA2.
doublearray MSCAL.NEWUVW() gives the 3-vector of UVW coordinates in J2000 in meters. It recalculates them, thus does not
return the UVW coordinates stored in the MeasurementSet.
doublearray MSCAL.NEWUVWWVL() gives the 3-vector of calculated UVW coordinates in J2000 in wavelengths for the reference
frequency of the appropriate spectral window.
doublearray MSCAL.NEWUVWWVLS() gives the nfreq*3-matrix of calculated UVW coordinates in J2000 in wavelengths for all channel
frequencies of the appropriate spectral window.
doublearray MSCAL.UVWWVL() gives the 3-vector of stored UVW coordinates in wavelengths for the reference frequency of the
appropriate spectral window.
doublearray MSCAL.UVWWVLS() gives the nfreq*3-matrix of stored UVW coordinates in wavelengths for all channel frequencies
of the appropriate spectral window.
By default all these functions will use the direction given in column PHASE_DIR of the FIELD subtable. It is
possible to use another column by giving its name as a string argument (e.g., HA(’DELAY_DIR’)).
Except for the last 2 functions, it is possible to use an explicit direction which must be given as [RA,DEC] in
J2000 or as a case-insensitive name of a planetary object (as deï¬ned by the Casacore Measures). For
example:
   derivedmscal.azel1([5h23m32.76, 10d15m56.49])
    derivedmscal.azel1(’MOON’)
The examples above give the azimuth and elevation of the given directions for each selected row in the
MeasurementSet, using the position of ANTENNA1 and the times in these rows.
If a string value is given, it is ï¬rst tried as a planetary object. Theoretically it is possible that a column
has the same name as a planetary object. In such a case the name can be escaped by a backslash to indicate
that a column name is meant. For example:
   derivedmscal.azel1(’\SUN’)
means that column SUN in the FIELD table has to be used.
Stokes conversion The STOKES function makes it possible to convert the Stokes parameters of a DATA column in a
MeasurementSet, for instance from linear or circular to iquv. It is also possible to convert the weights or
flags, i.e., to combine them in the same way as the data would be combined.
complexarray MSCAL.STOKES(complexarray, string) converts the data.
doublearray MSCAL.STOKES(doublearray, string) combines the weights.
boolarray MSCAL.STOKES(boolarray, string) combines the flags.
In all cases the case-insensitive string argument deï¬nes the output Stokes axes. It must be a comma separated
list of Stokes names. All values deï¬ned in the Casacore class Stokes are possible. Most important are:
XX, XY, YX, and/or YY. LINEAR or LIN means XX,XY,YX,YY.
RR, RL, LR, and/or LL. CIRCULAR or CIRC means RR,RL,LR,LL.
I, Q, U, and/or V. IQUV or STOKES means I,Q,U,V.
PTOTAL is the polarized intensity (sqrt(Q**2+U**2+V**2))
PLINEAR is the linearly polarized intensity (sqrt(Q**2+U**2))
PFTOTAL is the polarization fraction (Ptotal/I)
PFLINEAR is the linear polarization fraction (Plinear/I)
PANGLE is the linear polarization angle (0.5*arctan(U/Q)) (in radians)
If not given, the string argument defaults to ’IQUV’. For example:
  select mscal.stokes(DATA,’circ’) as CIRCDATA from my.ms
creates a table with column CIRCDATA containing the circular polarization data.
CASA style selection The BASELINE function makes it possible to do selection on baselines in a MeasurementSet or CalTable using
the special CASA selection syntax described in note 263. Similar functions CORR, TIME, FIELD, FEED,SCAN, SPW, UVDIST, STATE, OBS, and ARRAY can be used to do selection based on other meta data. The
functions accept a string containing a selection string and return a Bool value telling if a row matches the
selection string. For example,
  select from my.ms where mscal.baseline(’RT[2-4]’)
selects the cross-correlation baselines containing an antenna whose name matches the pattern in the function
argument. Note there is a difference how CASA and TaQL handle unknown antennas given in the baseline
selection string. CASA tasks give an error, while TaQL will not complain and not even report it,
because doing a selection this way should not behave differently from doing it like NAME=’RTX’.
Also note that in CASA tasks only one selection string per type can be given and the ï¬nal selection is the
AND of them. TaQL has the AND and OR operators making it possible to combine the selections in all kind
of ways, possibly using multiple selection strings of the same type.
Get values from a subtable Several functions exist to get information like the name of an antenna from the subtable for each row in the
main table. Basically they do a join of the main table and a subtable. For example:
  select mscal.ant1name(), mscal.ant2name() from my.ms
gets the names of the antennae used in each baseline.
The following functions can be used:
string MSCAL.ANT1NAME() gives the name of ANTENNA1.
string MSCAL.ANT2NAME() gives the name of ANTENNA2.
anytype MSCAL.ANT1COL(ColumnName) gives for ANTENNA1 the value in the given column (in quotes) in the ANTENNA subtable.
anytype MSCAL.ANT2COL(ColumnName) gives for ANTENNA2 the value in the given column (in quotes) in the ANTENNA subtable.
anytype MSCAL.STATECOL(ColumnName) gives for STATE_ID the value in the given column (in quotes) in the STATE subtable.
anytype MSCAL.OBSCOL(ColumnName) gives for OBSERVATION_ID the value in the given column (in quotes) in the OBSERVATION
subtable.
anytype MSCAL.SPWCOL(ColumnName) gives
for DATA_DESC_ID the value in the given column (in quotes) in the SPECTRAL_WINDOW
subtable.
anytype MSCAL.POLCOL(ColumnName) gives for DATA_DESC_ID the value in the given column (in quotes) in the POLARIZATION
subtable.
anytype MSCAL.FIELDCOL(ColumnName) gives for FIELD_ID the value in the given column (in quotes) in the FIELD subtable.
anytype MSCAL.PROCCOL(ColumnName) gives for PROCESSOR_ID the value in the given column (in quotes) in the PROCESSOR
subtable.
anytype MSCAL.SUBCOL(SubtableName, ColumnName, idcolumn) gives for the (integer) id-column the value in the given column in the given subtable. This is the
most common form and can be used to join any table with a subtable.
Note that the following are equivalent. The ï¬rst versions are shorthands for the latter ones.
 mscal.ant1name()
  mscal.ant1col(’NAME’)
  mscal.subcol(’ANTENNA’, ’NAME’, ANTENNA1)
In the last example the id-column must be given as such, thus must not be a string.
4.10.18 Special Measures functions
These functions make it possible to convert measures like directions, epochs, and positions from one
reference frame to another. All conversions supported by Casacore’s Measures are possible. For
example:
  meas.galactic (-6h52m36.7, 34d25m56.1, "J2000")
   meas.azel ("MOON", datetime(), "WSRT")
The ï¬rst example converts a J2000 position to galactic coordinates. The second example gives the moon’s
azimuth/elevation at the WSRT at the current date/time.
The following basic functions are available. Most functions return double angle values with unit rad. Only
the RISESET function returns date/time values. Note that all names used below are case-insensitive.
doublearray MEAS.DIR(toref, direction, epoch, position) converts a direction to the reference type given by the ’toref’ string. The epoch and position
arguments only need to be given if the conversion needs frame information (e.g., when converting
J2000 to apparent). Function name MEAS.DIRECTION can be used as well.
doublearray MEAS.DIRCOS(toref, direction, epoch, position) Same as function DIR, but returning 3 direction cosines instead of 2 angles. Function name MEAS.DIRECTIONCOSINE can be used as well.
DateTimearray MEAS.RISESET(direction, epoch, position) returns the rise and set date/times (UTC) of the sources given in the direction argument for the given dates and
positions. Function name MEAS.RISET can be used as well. Note that the source can be invisible all day (results
in setrise).
If visible all day, rise time is 0h0m and set time is 24 hours later. The TIME or CTIME function can be used on the result to get the time part only (as double or string).
If the sun is used as a source name (case-insensitive), it can be followed by a hyphen and one of the
following case-insensitive suffices indicating what to use. The default is CR which is used in most
almanacs.
CR: use the center of the sun with refraction correction.
UR: use the upper brim of the sun with refraction correction, thus show when part of the
sun is visible.
LR: use the lower brim of the sun with refraction correction, thus show when the full sun is
visible.
C: use the center of the sun without refraction correction.
U: use the upper brim of the sun without refraction correction.
L: use the lower brim of the sun without refraction correction.
CT: use the civil twilight (6 deg). without artiï¬cal light.
NT: use the nautical twilight (12 deg). visible for navigating.
AT: use the amateur astronomical twilight (15 deg).
ST: use the scientiï¬c astronomical twilight (18 deg).
The ï¬rst six suffices can also be used with the Moon. See stjarnhimlen.se for additional information.
doublearray MEAS.EPOCH(toref, epoch, position) converts an epoch to the reference type given by the ’toref’ string. The position argument only needs to
be given if the conversion needs frame information. By default conversions to sidereal time (e.g, LAST)
return the fraction giving the true sidereal time. Only if the toref string starts with ’F-’, ’F_’, ’f-’, or
’f_’ the full sidereal time is returned which includes the number of sidereal days since the start of
MJD.
doublearray MEAS.POS(toref, position) converts a position to the reference type given by the ’toref’ string. Function name MEAS.POSITION can be used as well.
The available reference types can be shown using the show meastypes command in the program taql. For ease of
use several specialized MEAS.DIR and MEAS.POS functions are deï¬ned with an implicit ’toref’
argument.
doublearray MEAS.J2000(direction, epoch, position) converts a direction to J2000.
doublearray MEAS.B1950(direction, epoch, position) converts a direction to B1950.
doublearray MEAS.APP(direction, epoch, position) converts a direction to apparent coordinates. Function name MEAS.APPARENT can be used as well.
doublearray MEAS.HADEC(direction, epoch, position) converts a direction to hourangle/declination.
doublearray MEAS.AZEL(direction, epoch, position) converts a direction to azimuth/elevation.
doublearray MEAS.ECL(direction, epoch, position) converts a direction to ecliptic coordinates. Function name MEAS.ECLIPTIC can be used as well.
doublearray MEAS.GAL(direction, epoch, position) converts a direction to galactic coordinates. Function name MEAS.GALACTIC can be used as well.
doublearray MEAS.SGAL(direction, epoch, position) converts a direction to supergalactic coordinates. Function name MEAS.SUPERGAL or MEAS.SUPERGALACTIC can be used as well.
doublearray MEAS.ITRFD(direction, epoch, position) converts a direction to ITRF coordinates. Function name MEAS.ITRFDIR or MEAS.ITRFDIRECTION can be used as well.
doublearray MEAS.LAST(epoch, position) converts an epoch to local sidereal time. Function name MEAS.LST can be used as well.
doublearray MEAS.ITRFxxx(position) converts a position to ITRF coordinates.
doublearray MEAS.WGSxxx(position) converts a position to WGS84 coordinates.
The names of the last two functions have a suffix xxx indicating how positions are returned. - XYZ means as x,y,z - LL or LONLAT means as lon,lat - H or HEIGHT means as height It defaults to XYZ.
The function arguments can be given in a variety of ways.
’toref’ is a constant scalar string giving the reference type to convert to. See the Measure classes
MDirection, MEpoch, and MPosition, for an overview of the types.
’direction’ gives one or more directions to convert. They can be given in several ways.
As a constant scalar or array of strings giving one or more planetary objects like MOON
or VENUS and/or giving the name of standard sources (CasA, CygA, TauA, VirA, HerA,
HydA, or PerA). In the future support for comets might be added. The names are case-insensitive.
As 2 constant double scalar arguments giving ra and dec (or longitude and latitude).
As a double array with an even number of elements giving ra/dec or longitude/latitude of
one or more directions. It can be a constant array (expression), but it can also be a column
or an expression using a column.
If a column or a column slice is given, the reference type stored in the column keywords will be
recognized. In other cases the input reference type should be given in the next string argument. If not
given, it defaults to J2000. For example:
  [’MOON’,’sun’, ’venus’]          # 3 planetary objects
   12h23m17.5, 23d56m43.8, ’B1950’  # ra/dec as scalar constants (as B1950)
   [12h23m17.5, 23d56m43.8]         # ra/dec as array (default J2000)
   PHASE_DIR[0,]                    # direction ra/dec in given column
’epoch’ gives one or more epochs to use. Similar to directions the reference type is taken from the
column keywords or can be given in the next argument. It defaults to UTC. Epochs can be given in three ways:
As a scalar or array containing double values. It can be a constant expression or a column
(expression).
As a scalar or array containing DateTime values.
As a scalar or array containing String values representing date/time. They will
automatically be converted to DateTime values using function datetime.
For example:
  datetime()                         # current date/time
   ’today’                            # current date/time
   [select unique TIME from my.ms]    # all times from some MS
   9Sep2011/12:00:00, ’UTC’           # given UTC time
Note that in the last example ’UTC’ is not necessary, because it is the default.
’position’ gives one or more directions to use. They can be given as x,y,z or as lon,lat with an optional
height. Usually the unit of the ï¬rst value deï¬nes if x,y,z or lon,lat is used. It is, however, also possible to
distinguish between LL and XYZ by using suffices like XYZ or LL in the reference type given in the
next argument.
As a scalar or array of observatory names using their positions in the Measures Observatory
table.
As 2 or 3 constant double scalar values giving xyz, lonlat, or lonlat/height.
As a double array giving one or more positions in xyz or lonlat. Similar to directions it can
be a column (expression) where the reference type is taken from the column keywords.
As two constant double arrays giving lonlat and height of one or more positions. The array
sizes have to match (thus the size of the lonlat array must be twice the size of the height
array).
If needed, the reference type (with optional suffix) can be given in the next argument. The reference
type defaults to ITRF if xyz coordinates are used, otherwise to WGS. For example:
  ’WSRT’                             # WSRT position
   5deg, 52deg                        # 2 scalar constants (WGS84 lonlat)
   (5deg, 52deg]                      # same, but as array
   5deg, 52deg, 5m                    # WGS84 lonlat with height
   [5deg, 52deg], [5m]                # same, but as array
   3.8288e+06m, 442449, 5.0649e+06    # xyz as scalars (ITRF)
   [41.84m, 4.835, 55.722], ’WGS’     # xyz as array (WGS84)
   POSITION                           # POSITION column
A few more elaborate examples are given below.
  meas.last (date(’15Oct2011/15:34’), 5deg, 52deg)
calculates the local apparent sidereal time for the given date and position.
  meas.azel ("JUPITER", [select unique TIME from ~/GER1.MS],
             ["WSRT","VLA"])
calculates Jupiter’s azimuth/elevation for WSRT and VLA for all times returned by the subquery (see next
section for subqueries).
  calc meas.b1950(PHASE_DIR[0,]) from ~/GER1.MS/FIELD’
converts the PHASE_DIR directions in the FIELD table to B1950. Note that no frame information is needed
for such a conversion.
  meas.azel([03h13m10,65d50m12], 24sep2015/12:0:0+[0:24]h, ’LOFAR’) deg
   meas.azel(03h13m10,65d50m12,’B1950’, 24sep2015/12:0:0+[0:24]h, ’LOFAR’)deg
calculates the azimuth/elevation (in degrees) of the given source direction for the LOFAR site for 24 hours.
The result is an array with shape [24,2]. The direction in the second example is given in B1950, the ï¬rst as
the default J2000.
4.11 Subqueries
As in SQL it is possible to create a set from a subquery. A subquery has the same syntax as a main query,
but has to be enclosed in square brackets or parentheses. Basically it looks like:
  SELECT FROM maintable WHERE time IN
       [SELECT time FROM othertable WHERE windspeed < 5]
The subquery on othertable results in a constant set containing the times for which the windspeed
matches. Subsequently the main query is executed and selects all rows from the main table with times in that
set. Note that like other bounded sets this set is transformed to a constant array, so it is possible to apply
functions to it (e.g., min, mean).
  SELECT [SELECT NAME FROM ::ANTENNA][ANTENNA1] FROM ~/GER1.MS
This example shows how a subquery is used to join the main table of a MeasurementSet and
its ANTENNA subtable. The subquery returns a list with the names of all antennae, which
subsequently is indexed with the antenna number to get the antenna name for each row in the main
table.
  SELECT mscal.ant1name() from ~/GER1.MS
is a newer and easier way to obtain the name of ANTENNA1. It makes use of the new user
deï¬ned functions in derivedmscal which can do an implicit join of a MeasurementSet and its
subtables.
  SELECT FROM maintable WHERE time IN
       [SELECT time FROM othertable WHERE windspeed <
            mean([SELECT windspeed FROM othertable])]
This example contains another subquery to get all windspeeds and to take the mean of them. So
the ï¬rst subquery selects all times where the windspeed is less than the average windspeed.
A subquery result should contain only one column, otherwise an exception is thrown.
It may happen that a subquery has to be executed twice because 2 columns from the other table are
needed. E.g.
  SELECT FROM maintable WHERE any(time >=
       [SELECT starttime FROM othertable WHERE windspeed < 5]
                                && time <=
       [SELECT endtime FROM othertable WHERE windspeed < 5])
In this case the other table contains the time range for each windspeed. For big tables it is expensive to
execute the subquery twice. A better solution is to store the result of the subquery in a temporary table and
reuse it.
  SELECT FROM othertable WHERE windspeed < 5 GIVING tmptab
   SELECT FROM maintable WHERE any(time >=
       [SELECT starttime FROM tmptab]
                                && time <=
       [SELECT endtime FROM tmptab])
However, this has the disadvantage that the table tmptab still exists after the query and has to be deleted
explictly by the user. Below a better solution for this problem is shown.
TaQL has a few extensions to support tables better, in particular the Casacore MeasurementSets.
The temporary problem above can be circumvented by using the ability to use a SELECT expression in
the FROM clause. E.g.
  SELECT FROM maintable,
       [SELECT FROM othertable WHERE windspeed < 5] tmptab
       WHERE any(time >= [SELECT starttime FROM tmptab]
              && time <= [SELECT endtime FROM tmptab])
However, below an even nicer solution is given.
The time range problem above can be solved elegantly by using a set as the result of the subquery.
Instead of a table name, it is possible to give an expression in the GIVING clause (as mentioned in
section 3.10). E.g.
  select from MY.MS where TIME in
       [select FROM OTHERTABLE where WINDSPEED < 5
            giving [TIME-INTERVAL/2 =:= TIME+INTERVAL/2]]
The set expression in the GIVING clause is ï¬lled with the results from the subquery and used in the
main query. So if the subquery results in 5 rows, the resulting set contains 5 intervals. Thereafter the
resulting intervals are sorted and combined where possible. In this way the minimum number of
intervals have to be examined by the main query.
In Casacore the other table will often be the name of a subtable, which is stored in a table or column
keyword of the main table. The standard keyword syntax can be used to indicate that the other table is
the table in the given keyword. Note that for a table keyword the :: part has to be given, otherwise the
name is treated as an ordinary table name. E.g.
  select from MY.MS where TIME in
       [select TIME from ::WEATHER where WINDSPEED < 5]
In this example the other table is a subtable of table my.ms. Its name is given by keyword WEATHER of
my.ms.
Often the result of a query on a subtable of a measurement set is used to select columns from the main
table. However, several subtables do not have an explicit key, but use the row number as an implicit
key. The function ROWID() can be used to return the row number as the subtable query result.
E.g.
  select from MY.MS where DATA_DESC_ID in
       [select from ::DATA_DESCRIPTION where
          SPECTRAL_WINDOW_ID in [0,2,4] giving [ROWID()]]
Note that the function ROWNUMBER cannot be used here, because it will give the row number in the
selection and not (as ROWID does) the row number in the original table. Furthermore, ROWID gives a
0-relative row number which is needed to be able to use it as a selection criterium on the 0-relative
values in the measurement set.
Select if any channel has a UV distance
100 wavelengths.
  select from MY.MS where any(sqrt(sumsqr(UVW[:2])) / c() *
       [select CHAN_FREQ from ::SPECTRAL_WINDOW][DATA_DESC_ID,]
          < 100)
In a MeasurementSet the UVW coordinates are stored in meters, so they have to be multiplied with
the frequency and divided by speed of light to get them in wavelengths. Because TaQL has no proper join operation, it is not possible to select directly on the
DATA_DESC_ID. However, using a nested query and indexing the result with the DATA_DESC_ID
has the same effect. It only requires that CHAN_FREQ has the same length in all rows in the subtable.
Using the new derivedmscal functions, below a much nicer solution is given.
  select from MY.MS where any(mscal.uvwwvls() < 100)
It shows how the UVWWVLS function in derivedmscal can be used to obtain the UVW coordinates
in wavelengths.
Calculate the angular distance between the Mars and Jupiter as seen from the WSRT for the coming 30
days.
  calc angdist(meas.app(’mars’,    date()+[0:31], ’WSRT’),
                meas.app(’jupiter’, date()+[0:31], ’WSRT’))
5 Aggregation, GROUPBY, HAVING
Similar to SQL it is possible to do aggregation and grouping in TaQL and to do selection on the groups using
the HAVING clause.
5.1 Aggregation and GROUPBY
One or more aggregated values can be calculated for a group deï¬ned by the GROUPBY clause. The
aggregate functions described in section 4.10.13 can be used. For example:
  SELECT ANTENNA1, ANTENNA2, gcount(), sqrt(sumsqr(UVW[:2]))
       FROM my.ms GROUPBY ANTENNA1,ANTENNA2
A group is formed for the unique values of the columns given in the GROUPBY clause. In the example
above a group per baseline is formed. Usually an aggregate function is ued to calculate a value for the group.
In the example above the aggregate function gcount() counts the number of rows per baseline.
Often only the GROUPBY columns and aggregated values are part of the SELECT clause, but the example
shows that other values (here the baseline length) can also be selected. Non-aggregated values get the values
in the last row of a group.
Usually aggregated values and GROUPBY are used jointly, but it is possible to leave out one of them. If
GROUPBY is not given, the entire table is a single group. For example:
  SELECT gcount() from my.ms
does not have groups, thus shows the total number of rows in the MS.
  SELECT ANTENNA1,ANTENNA2 from my.ms GROUPBY ANTENNA1,ANTENNA2
does not use aggregate functions, but shows the unique baselines in the MS. Apart from the order, it has the
same result as
  SELECT ANTENNA1,ANTENNA2 from my.ms ORDERBY UNIQUE ANTENNA1,ANTENNA2
but is somewhat faster.
In the examples above a sole aggregate function is used, but it is also possible to use it in an expression.
Similarly, an expression can be used in the GROUPBY. For example:
  select ctod(gmean(TIME)), gcount() from ~/data/GER.MS
     groupby round((TIME -
       [select gmin(TIME) from ~/data/GER.MS][0])/INTERVAL/5)
groups the MS in chunks of 5 time slots. Note that the nested query gets the TIME of the ï¬rst time slot.
The result is a set, hence the 0th element has to be taken.
Note that an aggregate function can only be used in the SELECT and HAVING clause, so TaQL will give
an error message if used elsewhere.
5.2 HAVING
The HAVING clause can be used to select speciï¬c groups. For example:
  SELECT TIME, gmax(amplitude(DATA)) as MAXA from my.ms GROUPBY TIME
       HAVING MAXA > 100
   SELECT TIME, gmax(amplitude(DATA)) from my.ms GROUPBY TIME
       HAVING gmax(amplitude(DATA)) > 100
groups by time, but only selects the groups for which the maximum amplitude of the DATA is more
than 100. Both examples give the same result, but the ï¬rst one is more efficient. Not only it is
less typing, but it is faster because it reuses the result column MAXA of the SELECT part.
Similar to WHERE, any expression can be used in HAVING, but the result has to be a bool scalar
value.
As shown in the example, HAVING will normally use aggregate functions, but it is not strictly needed.
However, selections without an aggregate function could as well be done in the WHERE clause.
Usually HAVING will be used in combination with GROUPBY, but it can be used without. It can
also be used without an aggregate function in the SELECT. However, it is an error if both are
omitted.
6 Some further remarks
6.1 Joining tables
As discussed in some previous sections it is possible to join tables on row number. Two examples show how to
do it.
6.1.1 Join on row number
  SELECT FROM mytable t1,othertable t2
     WHERE not all(t1.DATA ~= t2.DATA)
This command can be used to check if the data in mytable is about equal to the data in othertable. Both
tables have to have the same number of rows. The join is done on row number, thus the data in corresponding rows are compared.
6.1.2 Join using an indexed subquery
  SELECT [SELECT NAME FROM ::ANTENNA][ANTENNA1]
          FROM ~/GER1.MS
This example shows how a subquery is used to join the main table of a MeasurementSet with its
ANTENNA subtable. The subquery returns a list with the names of all antennae, which subsequently
is indexed with the antenna number to get the antenna name for each row in the main table.
The join is done using the ANTENNA1 column which gives the row number in the subtable, thus the index
in the subquery result.
6.1.3 Join using a subquery set
  SELECT FROM ~/GER1.MS WHERE ANTENNA1 IN
          [SELECT ROWID() FROM ::ANTENNA WHERE NAME ~ p/CS*/]
This example shows another way to use a subquery for a join of the main table of a MeasurementSet with its
ANTENNA subtable. It selects all baselines for which the ï¬rst station is a core station. The subquery returns
a set containing the ids of the core stations, which is used to select the correct stations in the main
table.
6.1.4 Join using derivedmscal
Several UDFs in the derivedmscal library make it possible to easily join a MeasuementSet or CASA
Calibration Table with a subtable like ANTENNA or SPECTRAL_WINDOW. These functions know which
columns to use making the join straightforward like in
  SELECT mscal.ant1name(), mscal.ant2name() from ~/GER1.MS
The library also contain the more general SUBCOL function making it possible to join any table with a
subtable. For example:
  SELECT mscal.subcol(’NAMES’,’NAME’,NAMEID) from obs.parmdb
to get the parameter name for a LOFAR ParmDB table. A ParmDB table has a subtable NAMES
containing the NAME and other info of a parameter. The column NAME_ID is used to reference that
subtable.
6.2 Optimization
A lot of development work could be done to improve the query optimization. At this stage only a few simple
optimizations are done.
Constant subexpressions are calculated only once. E.g. in COL*sin(180/pi()) the part sin(180/pi()) is evaluated once.
If a subquery generates intervals of reals or dates, overlapping intervals are combined and eliminated.
E.g.
  select from GER.MS where TIME in [select from ::POINTING where
    sumsqr(DIRECTION[1])>0 giving [TIME-INTERVAL/2=:=TIME+INTERVAL/2]]
can generate many identical or overlapping intervals. They are sorted and combined where possible to
make the set as small as possible.
If the righthand side of the IN operator is a single value, IN is turned into ==.
If the righthand side of the IN operator is a set of integer values with a min-max range of
=1024*1024,
that set is turned into a boolean vector to get linear lookup time.
TaQL does not recognize common subexpressions nor does it attempt to optimize the query. It means
that the user can optimize a query by specifying the expression carefully. When using operator
or &&,
attention should be paid to the contents of the left and right branches. Both operators evaluate the right
branch only if needed, so if possible the left branch should be the shortest one, i.e., the fastest to
evaluate.
The user should also use functions, operators, and subqueries in a careful way.
SQUARE(COL) is (much) faster than COL**2 or POW(COL,2), because SQUARE is faster. It is also
faster than COL*COL, because it accesses column COL only once. Similarly SQRT(COL) is faster than COL**0.5 or POW(COL,0.5)
SQUARE(U) + SQUARE(V) 1000**2 is considerably faster than SQRT(SQUARE(U) + SQUARE(V)) 1000, because the SQRT function does not need to be evaluated for each row.
TIME IN []
is faster than TIME0&& TIME4,
because in the ï¬rst way the column is accessed only once.
Returning a column from a subquery can be done directly or as a set. E.g.
  SELECT FROM maintable WHERE time IN
       [SELECT time FROM othertable WHERE windspeed < 5]
could also be expressed as
  SELECT FROM maintable WHERE time IN
       [SELECT FROM othertable WHERE windspeed < 5 GIVING [time]]
The latter (as a set) is slower. So, if possible, the column should be returned directly. This is also easier
to write. An even more important optimization for this query is writing it as:
  SELECT FROM maintable WHERE time IN
       [SELECT DISTINCT time FROM othertable WHERE windspeed < 5]
Using the DISTINCT qualiï¬er has the effect that duplicates are removed which often results in a much
smaller set.
Testing if a subquery contains at least N elements can be done in two ways:
  count([select column from table where expression]) >= N
 and
   exists (select from table where expression limit N)
The second form is by far the best, because in that case the subquery will stop the matching process as
soon as N matching rows are found. The ï¬rst form will do the subquery for the entire table.
Furthermore in the ï¬rst form a column has to be selected, which is not needed in the second
form.
Sometimes operator IN and function ANY can be used to test if an element in an array matches a value.
E.g.
  WHERE any(arraycolumn == value)
 and
   WHERE value IN arraycolumn
give the same result. Operator IN is faster because it stops when ï¬nding a match. If using ANY all
elements are compared ï¬rst and thereafter ANY tests the resulting bool array.
It was already shown in the section 4.8 that indexing arrays should be done with care.
7 Modifying a table
Usually TaQL will be used to get a subset from a table. However, as described in the ï¬rst sections, it
can also be used to change the contents of a table using the UPDATE, INSERT, or DELETE
command. Note that a table has to be writable, otherwise those commands exit with an error
message.
7.1 UPDATE
  UPDATE table SET update_list [FROM table_list]
                                [WHERE ...] [ORDERBY ...]
                                [LIMIT ...] [OFFSET ...]
updates all or some rows in the ï¬rst table. More input tables can be given in the FROM clause and used in
clauses like SET and WHERE. Unlike SQL it is possible to specify more tables in the UPDATE part which is
the same as specifying them in the FROM clause. However, using the FROM clause makes it more clear that
only the ï¬rst table is updated. update_list is a comma-separated list of column=expression parts. Each part tells to update the given
column using the expression. Both scalar and array columns are supported. E.g.
  UPDATE vla.ms SET ANTENNA1=ANTENNA1-1, ANTENNA2=ANTENNA2-1
to make the antenna numbers zero-based if accidently they were written one-based.
  UPDATE this.ms set DATA=t2.DATA, FLAG=t2.FLAG
              FROM that.ms t2 where all(FLAG)
   UPDATE this.ms, that.ms t2 set DATA=t2.DATA, FLAG=t2.FLAG
                              where all(FLAG)
are equivalent. They copy the DATA and FLAG column of that.ms to this.ms for rows where all data in
this.ms are flagged. Note the use of the shorthand (alias) t2.
If an array gets an array value, the shape of the array can be changed (provided it is allowed for that
table column). Arrays can also be updated with a scalar value causing all elements in the array to be set to
that scalar value.
  UPDATE vla.ms SET FLAG=F
It sets all elements of the arrays in column FLAG to False.
Type promotion and demotion will be done where possible. For example, an integer column can get the
value of a double expression (the result will be truncated). Unit conversion will be done as needed. Thus if a column and its expression have different units, the
expression result is automatically converted to the column’s unit. Of course, the units must be of the same
type to be able to convert the data.
Note that if multiple column=expression parts are given, the columns are changed in the order as
speciï¬ed in the update-list. It means that if an updated column is used in an expression for a later column,
the new value is used when evaluating the expression. e.g., in
  UPDATE vla.ms SET DATA=DATA+1, SUMD=sum(DATA)
the SUMD update uses the new DATA values.
Thus to swap the values of the ANTENNA1 and ANTENNA2 column, one can not do:
  UPDATE vla.ms SET ANTENNA1=ANTENNA2, ANTENNA2=ANTENNA1
To solve this problem a temporary table (in this case in memory) can be used to save the value of e.g.,
ANTENNA1:
  UPDATE my.ms
       set ANTENNA1 = ANTENNA2, ANTENNA2 = orig.ANTENNA1
       FROM [select ANTENNA1 from my.ms giving as memory] orig
  UPDATE vla.ms SET FLAG[1,1]=T
   UPDATE vla.ms SET FLAG[1,]=T
The ï¬rst example sets only a single array element, while the second one sets an entire row in the array.
Similar to numpy it is also possible to use a mask like
  UPDATE vla.ms SET FLAG[isnan(DATA)]=T
which sets the flag for the DATA values being a NaN. The data and mask must have the same shape. Note
this is easier to write than the similar command
  UPDATE vla.ms SET FLAG = iif(isnan(DATA), T, FLAG)
Masking and slicing can be combined making it possible to use masking on a part of an array. If the mask is
given ï¬rst, the slice is taken from both the data and mask. If the slice is given ï¬rst, it is only applied to the
data; the mask should have the same shape as the slice. For example:
  UPDATE vla.ms SET FLAG[isnan(DATA)][,0]=T
   UPDATE vla.ms SET FLAG[,0][isnan(DATA[,0])]=T
Both commands set the flag for NaN data in the XX polarization. The ï¬rst one is somewhat easier to write,
but processes the entire DATA and FLAG before taking the slice. The second one only reads and processes
the required parts of DATA and FLAG, thus is more efficient.
7.1.2 Update columns from a masked array
If a column is updated with the value of a masked array, only the array part of the masked array is used.
However, it is also possible to jointly update the data column and mask column from a masked array by
combining them in parentheses like:
  UPDATE vla.ms SET (DATA,FLAG)=maskedarray
It writes the data part into DATA and the mask into FLAG. As above it is possible to use a slice or mask
operator on the combination like:
  UPDATE vla.ms SET (DATA,FLAG)[,0]=maskedarray
   UPDATE vla.ms SET (DATA,FLAG)[isnan(DATA)]=maskedarray
The slice or mask is applied to both columns.
7.2 INSERT
The INSERT command adds rows to the table. It can take three forms:
  INSERT INTO table_list SET column=expr, column=expr, ...
   INSERT INTO table_list [(column_list)] VALUES (exprlist),(exprlist), ...
   INSERT INTO table_list [(column_list)] SELECT_command
The ï¬rst and second form are basically equivalent, but differ in syntax. The ï¬rst form has the same syntax
as the UPDATE command, while the second form is the SQL syntax making it possible to leave out the
column names (see below). In both forms it is possible to jointly specify data column and mask column if the
value is a masked array. This is done by combining them in parentheses like (DATA,FLAG) as described in the
previous subsection for the UPDATE command.
The ï¬rst form adds one row to the table and puts the values given in the expressions into the columns.
For example:
  INSERT INTO my.ms SET ANTENNA1=0, ANTENNA2=1
adds one row, puts 0 in ANTENNA1 and 1 in ANTENNA2.
The second form can add multiple rows to the table. It puts the values given in the expression lists into
the columns given in the column list. If the column list is not given, it defaults to all stored columns in the
table in the order as they appear in the table description. Multiple expression lists can be given; each list
results in the addition of a row. Each expression in the expression list can be as complex as needed; for
example, a subquery can also be given. Note that a subquery is evaluated before the new row is added,
so the new row is not taken into account if the subquery is done on the table being modiï¬ed.
It should be clear that the number of columns has to match the number of expressions. Note that row cells not mentioned in the column list, are not written, thus may contain rubbish in the new
rows. The data types and units of expressions and columns have to conform in the same way as for the UPDATE
command; values have to be convertible to the column data type and unit. For example:
  INSERT INTO my.ms (ANTENNA1,ANTENNA2) VALUES (0,1),(2,3)
adds two rows, putting 0 and 2 in ANTENNA1 and 1 and 3 in ANTENNA2.
The LIMIT clause can be used to add multiple rows while giving fewer expressions. LIMIT can be given
at the beginning or the end of the command. For example:
  INSERT INTO my.ms (COL1) VALUES (rowid()) LIMIT 100
   INSERT LIMIT 5 INTO my.ms (COL1,COL2) VALUES (0,0),(1,1)
The ï¬rst example will add 100 rows where the value in each row is the row number. The second example
shows that multiple expression lists can be given. It will iterate through them while adding rows. Thus COL1
and COL2 will have the values 0, 1, 0, 1, and 0 in the new rows.
The third form evaluates the SELECT command and copies the rows found in the selection to the table
being modiï¬ed (which is given in the INTO part). The columns used in the modiï¬ed table are deï¬ned in the
column list. As above, they default to all stored columns. The columns used in the selection have to be
deï¬ned in the column-list part of the SELECT command. They also default to all stored columns.
For example:
  INSERT INTO my.ms select FROM my.ms
appends all rows and columns of my.ms to itself. Please note that only the original number of rows is
copied.
  INSERT INTO my.ms (ANTENNA1,ANTENNA2) select ANTENNA2,ANTENNA1
    FROM other.ms WHERE ANTENNA1>0
copies rows from other.ms where ANTENNA10.
It swaps the values of ANTENNA1 and ANTENNA2. All other columns are not written, thus may contain
rubbish.
7.3 DELETE
  DELETE FROM table_list
     [WHERE ...] [ORDERBY ...] [LIMIT ...] [OFFSET ...]
deletes some or all rows from a table.
  DELETE FROM my.ms WHERE ANTENNA1>13 OR ANTENNA2>13
deletes the rows matching the WHERE expression. If no selection is done, all rows will be deleted. It is possible to specify more than one table in the FROM clause to be able to use, for example,
keywords from other tables. Rows will be deleted from the ï¬rst table mentioned in the FROM
part.
8 Creating a new table
TaQL can be used to create a new table. The data managers to be used can be given in full detail. The
syntax is:
  CREATE TABLE tablename AS options colspecs LIMIT nrows DMINFO datamanagers
The command consists of 4 parts, all of them optional.
The table name and options can be given in the same way as in the GIVING clause.
The columns are deï¬ned in the colspecs part. If not given, a table without any column is
created. Below column speciï¬cation is described in more detail.
An expression giving the number of rows can be speciï¬ed in the LIMIT part. If not given, it
defaults to 0.
For expert users data managers can be deï¬ned in the optional DMINFO part described further
down.
The CREATE TABLE command can be used in a nested query making it possible to ï¬ll it immediately.
For example:
  update [create table a.tab col1 int limit 10] set col1=rowid()
creates a table with one column and ten rows. The column is ï¬lled with the row number. Note that the
following command would do the same.
  select rowid() as col1 int limit 10 giving a.tab
8.1 Column speciï¬cation
The colspecs part deï¬nes the column names, their data types, and optional shapes and units. It can
optionally be enclosed in square brackets or parentheses (for SQL compatibility). It is a comma separated list
of column speciï¬cations. Each speciï¬cation looks like:
  columnname datatype [NDIM=n, SHAPE=[d1,d2,...], UNIT=’s’,
                        DMTYPE=’s’, DMGROUP=’s’, COMMENT=’s’]
The possible data type strings are given in section 4.1. The part enclosed in square brackets is optional. Zero
or more of these keywords can be used. It makes it possible to deï¬ne array columns and/or default
data manager to be used. The square brackets are optional if only one such keyword is used.
NDIM=n deï¬nes if the column contains scalars or arrays. A negative value means a scalar, which is the default (unless shape is also given). A value 0 means
an array of any dimensionality. A positive value means an array with the given dimensionality.
SHAPE=[d1,d2,...] makes it possible to deï¬ne the exact array shape. If given and if NDIM is positive, they should be consistent.
UNIT=’s’ deï¬nes the unit to be used for the column. It can be any valid unit (simple or compound). It is a string, thus must always be enclosed in
quotes.
COMMENT deï¬nes comments for the column. It has a string value, thus quotes have to be used.
DMTYPE, DMGROUP are rather speciï¬c and are for the expert user. They have a string value, thus quotes have to be used.
8.2 Data manager speciï¬cation
The datamanagers part makes it possible for the expert user to deï¬ne the data managers to be used by
columns. It is a comma separated list of data manager speciï¬cations looking like the output of the
table.getdminfo command in Python. Each speciï¬cation has to be enclosed in square brackets. For
example:
  dminfo [NAME="ISM1",TYPE="IncrementalStMan",COLUMNS=["col1"]],
          [NAME="SSM1",TYPE="StandardStMan",
           SPEC=[BUCKETSIZE=1000],COLUMNS=["col2","col3"]]
The case of the keyword names used (e.g., NAME) is important. They have to be given in uppercase. The
following keywords can be given: NAME deï¬nes the unique name of the data manager. TYPE deï¬nes the type of data manager. SPEC is a list of keywords giving the characteristics of the data manager. This is highly data manager type
speciï¬c. If shapes have to be given here, they always have to be in Casacore format, thus in Fortran order.
TaQL has no knowledge about these internals. COLUMNS is a list of column names deï¬ning all columns that have to be bound to the data manager.
9 Modifying the table structure
TaQL can be used to modify the table structure, i.e., to add, rename, and remove columns and keywords. It
is also possible to add rows. The syntax is:
  ALTER TABLE tablename FROM table_list subcommand_list
It changes the table with the given name. The tables given in the optional FROM clause can be used in
expressions deï¬ning keyword values. Any number of subcommands can be given, separated by
whitespace and/or comma. The following subcommands can be given. They are explained in the next
subsections.
  ADD COLUMN colspecs DMINFO datamanagers
   RENAME COLUMN old TO new, old TO new, ...
   DELETE COLUMN column_list
   SET KEYWORD name=value AS dtype, ...
   COPY KEYWORD name=name AS dtype, ...
   RENAME KEYWORD old TO new, old TO new
   DELETE KEYWORD keyword_list
   ADD ROW nrows
The nouns COLUMN and KEYWORD can also be given in the plural form. The whitespace
between verb and noun is optional. For SQL-compatibility DROP can be used instead of DELETE.
For example:
  ALTER TABLE my.tab RENAME COLUMN Col1 to Col1A, ADDCOLUMNS Col1 I4
renames column Col1 to Col1A and adds a new column Col1 with data type I4.
Note that TaQL has no way of showing keywords that have a record value. The program showtable can be
used for that purpose.
9.1 ADD COLUMN
  ADD COLUMN colspec DMINFO datamanagers
adds one or more columns to the table. The speciï¬cation of the columns and the optional data managers is
the same as used in the CREATE TABLE command. Thus for each column a data type, dimensionality or
shape, and unit can be given. The data manager(s) for the new columns can be speciï¬ed in the DMINFO
part. If not given, StandardStMan will be used. For example:
  ADD COLUMN NCol1 R4, NCol2 R8 [UNIT="m", NDIM=3]
adds two columns, a 4-byte floating point scalar column and an 8-byte floating point 3-dim array column.
They will be stored with StandardStMan.
9.2 RENAME COLUMN
  RENAME COLUMN old1 TO new1, old2 TO new2, etc.
renames one or more columns in a table. For example:
  RENAME COLUMN NAME to NAME_SAV, ADDR to ADDR_SAV
9.3 DELETE COLUMN
  DELETE COLUMN col1, col2, etc.
removes one or more columns. Note that if multiple columns are combined in a TiledStMan, they have to be
removed at the same time. Thus in that case
  DELETE COLUMN col1, col2
   DELETE COLUMN col1, DELETE COLUMN col2
are not the same, because the second example might fail.
9.4 SET KEYWORD
  SET KEYWORD key1=value1 AS dtype, etc.
adds a keyword with the given value or replaces the value if the keyword already exists. The value of a
keyword can be a scalar, array, or arbitrarily deeply nested record. See section 4.5 how to specify a keyword
name in a column or nested record. The AS dtype part can be used to explicitly set the data type of a new
keyword. For an existing keyword, the data type of the new value has to match the data type of the current
value.
The value can be an expression, possibly using values from another table given in the FROM clause. It
has to be a constant expression, thus cannot depend on column values. Of course, column values can be used
when aggregated to a single value. If no data type is given, the data type of the expression result is used. If
given, upward and downward coercion is possible (e.g., integer to float and also float to integer). For
example:
  SET KEYWORD key1=4
   SET KEYWORD ::key1=4+5 AS U4
   SET KEYWORD key1 = otherkey as I4
   SET KEYWORD col::ckey.subrec.fld1 = [4,5,6.]
   SET KEYWORD col::ckey=[=], col::ckey.subrec=[=]
   SET KEWYORD key=[] AS I4
The 1st example sets table keyword key1 to 4. Its data type is not given, thus is the expression’s data type,
in this case I8. The 2nd example sets key1 to 9, but as an unsigned 4 byte integer. Note that the :: part is redundant.
The 3rd example copies the value of keyword otherkey while converting its data type to I4. Note that if no
data type is given, the data type of otherkey is NOT preserved, because it is seen as a TaQL expression which
has data type I8 (or R8). The 4th example sets the ckey.subrec.fld1 in column col to the given vector. It is a nested structure, thus ï¬eld
fld1 in ï¬eld subrec of column keyword ckey will be set. Its data type will be R8. Note that the command in the 4th example does not create the higher level records. If not existing yet, the
5th example can be used to create them, where [=] denotes an empty record (it is the old Glish syntax for an
empty struct). The last example shows how to create a key with an empty integer vector as value. In such a case the data
type must be given, because it cannot be derived from the value.
Setting a keyword to the value of another keyword is easily possible. For instance:
  SET KEYWORD key2 = otherkey
However, it has two problems. 1) As explained above the data type might not be preserved. 2) Keywords having a record value cannot be copied this way, because TaQL expressions do not support
record values.
9.5 COPY KEYWORD
  COPY KEYWORD key = otherkey AS dtype, etc.
copies the value of keyword otherkey to key. It can be used for any keyword value, thus also for records. The
optional AS dtype part can be used to change the data type.
9.6 RENAME KEYWORD
  RENAME KEYWORD old1 TO new1, old2 TO new2, etc.
renames one or more table or column keywords. If the old keyword is a ï¬eld in a column or a nested
record, the new name should only contain the new ï¬eld name, not the full keyword path. For
example:
  RENAME KEYWORD NAME to NAME_SAV, Col1::CNAME to CNAME_SAV
   RENAME KEYWORD KEYS.SET.NAME to NAME_SAV
The ï¬rst example renames the table keyword NAME and the keyword CNAME of column Col1.
The second example renames a ï¬eld in the nested records of table keyword KEYS.
9.7 DELETE KEYWORD
  DELETE KEYWORD key1, key2, ...
removes one or more table or column keywords.
9.8 ADD ROW
adds the given number of rows to the table.
  ADD ROW nrows
where nrows can be any expression. For example,
 ALTER TABLE mytab ADD ROW [SELECT GCOUNT() from othertab]
makes mytab the same size as othertab (assuming it was empty).
10 Counting in a table
Before TaQL had the GROUPBY command, the COUNT command could be used instead of the gcount
aggregate function to count the number of occurrences in a table. For backward compatibility this command can still be used, but its usage is discouraged, also because usually
GROUPBY is faster.
The exact syntax is:
  COUNT column-list FROM table-list [WHERE expression]
It counts the number of rows for each unique tuple in the column list of the table (after the possible
WHERE selection is done). For example:
  COUNT TIME FROM my.ms
counts the number of rows per timestamp.
  COUNT ANTENNA1,ANTENNA2 FROM my.ms
counts the number of rows per baseline.
As in the other TaQL commands a column in the column list can be any expression, but that will be
slower than straight columns.
11 Calculations on a table
TaQL can be used to get derived values from a table by means of an expression. The expression can
result in any data type and value type. For example, if the expression uses an array column, the
result might be a vector of arrays (an array for each row). If the expression uses a scalar column,
the result might be a vector of scalars or even a single scalar if a reduce function like SUM is
used.
The CALC command was developed before the GROUPBY was available and before SELECT could be
used without the FROM part. Currently, SELECT is more powerful than the CALC command. For example,
multiple expressions can be given in a SELECT command. However, especially in Python sessions CALC
has the advantage that it returns the results as a numpy-array or a list instead of a Casacore
table.
The exact syntax is:
  CALC expression [FROM table_list]
The part in square brackets can be omitted if no column is (directly) used in the expression. The examples
will make clear what that means. The following syntax is still available for backward compatibility:
  CALC FROM table_list CALC expression
  CALC 1in cm
is a simple expression not using a table. It shows how the CALC command can be used as a desk calculator
to convert 1 inch to cm.
  CALC mean(column1+column2) FROM mytable
gives a vector of scalars containing the mean per row.
  CALC sum([SELECT FROM mytable GIVING [mean(column1+column2)]])
gives a single scalar giving the sum of the means in each row. Note that in this command the CALC
command does not need the FROM clause, because it does not use a column itself. Columns are only used in
the nested query which has a FROM clause itself.
12 Examples
12.1 Selection examples
Some examples are given starting with simple ones.
12.1.1 Reference table results
The result of the following queries is a reference table, because no expressions have been given in the
column-list. This will be the most common case when using TaQL.
SELECT FROM some.ms WHERE ANTENNA1 != ANTENNA2 selects the cross-correlations in a MeasurementSet.
SELECT NAME FROM some.ms::ANTENNA selects the NAME of all antennae in a MeasurementSet.
SELECT unique ANTENNA1,ANTENNA2 FROM some.ms gives the baselines used in a MeasurementSet.
SELECT ANTENNA1,ANTENNA2 FROM some.ms GROUPBY ANTENNA1,ANTENNA2 does the same using the GROUPBY clause.
SELECT FROM mytable ORDERBY column0 DESC limit 10 selects the 10 highest values of column0.
SELECT FROM some.MS WHERE near(MJD(1999/03/30/17:27:15), TIME) selects the rows with the given time from a MeasurementSet. Note that the TIME is stored in seconds, but will automatically be converted to days.
SELECT FROM some.MS where TIME in     [{MJD(1999/03/30/17:27:15),MJD(1999/03/30/17:29:15)}] selects the rows in the given closed time interval.
SELECT FROM some.MS where TIME in     [MJD(1999/03/30/17:27:15),MJD(1999/03/30/17:29:15)] selects the rows having one of the given times. Note the difference with the previous example where an interval was given. Here a set of two
individual time values is given.
SELECT NAME FROM some.ms::ANTENNA WHERE NAME ! p/[CR]S*/ selects the names of the international LOFAR stations (not core or remote).
SELECT FROM some.ms WHERE ntrue(FLAG) >= 3 selects rows where at least 3 visibilities are flagged.
SELECT FROM book.table WHERE nelements(author) > 1 selects books with more than 1 author.
SELECT FROM some.ms WHERE any(ANTENNA1==[0,0,1] && ANTENNA2==[1,3,2]) selects the antenna pairs (baselines) 0-1, 0-3, and 1-2. It requires some explanation. The two comparisons result in boolean vectors (with 3 elements).
It matching elements are both true, the baseline in the table row matches. Thus the vectors are
and-ed to see if any two matching elements are true.
SELECT FROM some.ms WHERE ANTENNA1 in [0,0,1] && ANTENNA2 in [1,3,2]) looks the same as above, but will select all baselines between the two sets, thus also 1-1, 1-3,
and 0-2.
SELECT FROM some.ms t1, that.ms t2 WHERE !all(near(t1.DATA, t2.DATA, 1e-5)) selects all rows where the DATA columns in both tables are not equal (with some tolerance).
Note the use of shorthands t1 and t2.
SELECT FROM mytable WHERE cos(0d1m) <= Â Â Â Â sin(52deg) * sin(DEC) + cos(52deg) * cos(DEC) * cos(3h30m - RA) selects observations with an direction (in say J2000) inside a cone with a radius of 1 arcmin
around (3h30m, 52deg). To ï¬nd them the condition DISTANCERADIUS
must be fulï¬lled, which is equivalent to COS(RADIUS)COS(DISTANCE).
SELECT FROM mytable WHERE [RA,DEC] INCONE [3h30m, 52deg, 0d1m] does the same as above in an easier (and faster) way.
SELECT FROM mytable WHERE angdist([RA,DEC], [3h30m, 52deg]) <= 0d1m] is another way to do the above.
SELECT FROM mytable WHERE object == pattern("3C*") &&     [RA,DEC] INCONE [3h30m, 52deg, 0d1m] ï¬nds all 3C objects inside that cone.
SELECT ANTENNA1,ANTENNA2,sqrt(sumsqr(UVW[:2]))     FROM some.ms GROUPBY ANTENNA1,ANTENNA2 ï¬nds the 10 longest baselines. It groups by ANTENNA1 and ANTENNA2 to get the unique
baselines. UVW[:2] denotes the U and V coordinate giving the baseline length.
select from MY.MS where DATA_DESC_ID in [select from ::DATA_DESCRIPTION where     SPECTRAL_WINDOW_ID in [0,2,4] giving [ROWID()]] ï¬nds all rows in a measurement set matching the given spectral windows. It uses a nested query
to ï¬nd the DATA_DESC_ID for each spectral window.
select from MY.MS where TIME in [select from ::SOURCE where REST_FREQUENCY <180MHz     giving [TIME-INTERVAL/2 =:= TIME+INTERVAL/2]] ï¬nds all rows in a measurement set observing sources with a rest frequency less than 180 Mhz.
select from VLA.MS,     [select from VLA.MS where sumsqr(UVW[:2]) < 625] as TIMESEL     where TIME in [select distinct TIME from TIMESEL]     && any([ANTENNA1,ANTENNA2] in [select from TIMESEL giving     [iif(UVW[2] < 0, ANTENNA1, ANTENNA2)]]) selects rows where an antenna (VLA has 25 m diameter) is shadowed. The query in the FROM command ï¬nds all rows where an antenna is shadowed (i.e., its
UV-distance less than 25 meters) and creates a temporary table. This selection (named
TIMESEL) is done ï¬rst otherwise two 2 equal selections are needed in the main WHERE
command. The last line determines which antenna is shadowed (based on the W coordinate).
The two lines above selects the times and baselines where an antenna is shadowed.
select from MS     where DATA_DESC_ID in [select from ::DATA_DESCRIPTION     where SPECTRAL_WINDOW_ID in [select from ::SPECTRAL_WINDOW     where NET_SIDEBAND==1 giving [ROWID()]] giving [ROWID()]] ï¬nds all rows in the MeasurementSet with the given NET_SIDEBAND. The MeasurementSet uses a table to map spectral-window-id to data-desc-id. Hence two nested
subqueries are needed.
select findcone(REFERENCE_DIR[0,],     [16h34m33.805,62d45m36.83, 12h29m06.7,2d3m9], 1arcsec)     from MS/FIELD compares the direction given in the ï¬rst argument with the directions given in the second function
argument using the search radius given in the third argument. It returns the index of the ï¬rst
matching cone (thus 0 or 1). If no cone matches, it returns -1. It can be used in the following example to ï¬nd the name of the source matching a direction.
select [’unknown’,’3C343’,’3C273’][1+findcone(...)] from MS/FIELD where ... is the
ï¬ndcone argument list given in the previous example. 1 is added to cope with the case that no
cone matches.
12.1.2 Plain table results
The following examples result in a plain table, thus in a deep copy of the query results, because the
column-list contains an expression or a data type.
SELECT column0+column1 FROM mytable creates a table of 1 column with name Col_1. Its data type is on the expression data type.
SELECT column0+column1 Res I4 FROM mytable creates a table of 1 column with name Res. Its data type is a 4 byte signed integer.
SELECT colx colx R4 FROM mytable creates a table of 1 column with name colx. The sole purpose of this selection is to convert the
data type of the column.
SELECT means(DATA,0) AS DATA_MEAN C4 FROM my.ms creates a table of 1 column with name DATA_MEAN. Column DATA in a Casacore MeasurementSet
is a 2-dimensional array with axes polarization and frequency. This command calculates and
stores the mean in each polarization. If no data type was given, the means would have been
stored as double precision complex (which is the expression data type). Note that this command is valid when using python style; in glish style MEANS(DATA,2) should
be used.
12.2 Modiï¬cation examples
update MY.MS set VIDEO_POINT=MEANS(DATA,2) where isdefined(DATA) sets the VIDEO_POINT of each correlation to the mean of the DATA for that correlation. Note
that the 2 indicates averaging over the second axis, thus the frequency axis.
update MY.MS set FLAG_ROW=T where isdefined(FLAG) && all(FLAG) sets FLAG_ROW in the rows where the entire FLAG array is set.
delete from MY.MS where FLAG_ROW deletes all flagged rows.
insert into MY.MS select from OTHER.MS where !FLAG_ROW copies all unflagged rows from OTHER.MS to MY.MS.
insert into MY.MS/DATA_DESCRIPTION Â Â Â Â (SPECTRAL_WINDOW_ID,POLARIZATION_ID,FLAG_ROW) Â Â Â Â values (1,0,F) adds a row to the DATA_DESCRIPTION subtable and initializes it.
12.2.1 Applying running median to an image
The following command shows how a running median can be applied to an Casacore image.
  update my.imgd set map = map - runningmedian(map,25,25)’)
The running medians are subtracted from the data in the copy. It uses a half window size of 25x25, thus the
full window is 51x51. When doing this, one should take care that in case of a spectral line cube the image is not too large,
otherwise it won’t ï¬t in memory. If too large, it should be done in chunks like:
  update my.imgd set map[,,sc:ec,] =
              map[,,sc:ec,] - runningmedian(map[,,sc:ec,],25,25)’)
where sc and ec are the start and end frequency channel. In this example it is assumed that the axes of the
image are RA, DEC, freq, Stokes. Note that the image is updated, so it should have been copied before if the original data needs to be
kept.
create table mytab colarr R4 ndim=0 creates a table of 1 array column with arbitrary dimensionality.
create table mytab colarr R4 [shape=[4,128], dmtype=’TiledColumnStMan’] creates a table of 1 array column with the given shape. The column is stored with the
TiledColumnStMan storage manager using its default settings.
create table mytab colarr R4 shape=[4,128] Â Â Â Â dminfo [TYPE=’TiledColumnStMan’, NAME=’TCSM’, Â Â Â Â SPEC=[DEFAULTTILESHAPE=[4,32,64]], COLUMNS=[’colarr’]] creates a table of 1 array column with the given shape. The column is stored with the
TiledColumnStMan storage manager using the given settings.
12.4 Calculation examples
calc 1+2 uses TaQL as a desktop calculator.
calc 7-Apr-2007 - 20-Nov-1979 calculates the number of days between these dates.
calc sum([select from MY.MS giving [ntrue(FLAG)]]) determines the total number of flags set in the measurement set.
calc mean(abs(DATA)) from [select from MY.MS where ANTENNA1==0]
calculates for each row the mean of the data for the selected subset of the measurement set.
calc mean([select from MY.MS where ANTENNA1==0 Â Â Â Â giving [mean(abs(DATA))]])
looks like the previous example. It, however, calculates the mean of the mean of the data in each
row for the selected subset of the measurement set.
calc max([select from MY.MS where isdefined(DATA) Â Â Â Â giving [max(abs(VIDEO_POINT-MEANS(DATA,0)))]])
shows the maximum absolute difference between VIDEO_POINT of each correlation and the
mean of the DATA for that correlation. Note that the 2 indicates averaging over the ï¬rst axis,
thus the frequency axis.
12.5 Aggregation/groupby examples
select gcount(*) from my.ms counts the number of rows in the table.
select TIME, gcount(*) from my.ms groupby TIME counts the number of rows (usually number of baselines) per time slot.
select ANTENNA1,ANTENNA2,gfirst(TIME),glast(TIME),gcount()     from my.ms groupby ANTENNA1,ANTENNA2 counts the number of rows (usually number of time slots) and shows the ï¬rst and last time per
baseline.
select gmean(DATA) from my.ms     groupby int((TIME - [select TIME from my.ms limit1][0]) / INTERVAL / 10)) calculates the average of DATA for every 10 time slots. Note it also averages in frequency
and polarization. The following example shows how to average each frequency channel and
polarization.
select boxedmean(gaggr(DATA), 10, 4) from my.ms     groupby int((TIME - [select TIME from my.ms limit1][0]) / INTERVAL / 10)) calculates the average of DATA per polarization for every 10 time slots and 4 frequency channels.
Note it it ï¬rst combines the data of each 10 time slots in a single array, after which the boxedmean
function is used to average every [10,4,1] box.
12.5.1 Obtaining the flux density from visibility data
The Miriad program uvflux estimates the source I flux density and its standard deviation at the phase
center without having to make an image. A single, not too complicated TaQL command (courtesy
Dijkema, Heald) provides the same functionality on a MeasurementSet. For LOFAR it is best to use
baselines with a length between 5 and 10 km. The command shows various aspects of TaQL that
are explained below. The numbers at the beginning of the lines point to the text following the
example.
4.    select gstddev(SUMMED) as STDVALS,
 4.           gmean(SUMMED) as MEANVALS,
 4.           gcount(SUMMED) as NVALS
 3.    from (select gmean(
 1.                  sum(iif(FLAG[,0:4:3], 0, abs(DATA[,0:4:3])))
 1.                   / nfalse(FLAG[,::3])
 3.                 ) as SUMMED
             from ~/data/GER.MS
 2.          where mscal.baseline(’5km~10km) && !all(FLAG)
 3.          groupby TIME)
A subquery is used to get the average flux (I = 0.5*(XX+YY)) per time slot.
For each baseline it gets the mean of the channels. Note that it uses sum/n to ignore flagged
visibilities. The iif function tells to use 0 for them. Also note that XX is the 1st and YY
the 4th polarisation, hence [0:4:3] (or [::3]) indexes these polarisations. Once masked arrays are
supported by TaQL, it could be written as: mean(DATA[,::3][FLAG[,::3]])
It only uses the baselines with lengths between 5 and 10 km where not all visibilities are flagged.
Note that a mscal user deï¬ned function is used for the baseline selection as described in section
4.10.17.
Thereafter the average flux per time slot is determined in the subquery using the gmean
aggregation and GROUPBY functionality. The result is an intermediate table with one column
called SUMMED and a row per time slot.
Finally, the outer query uses aggregate functions to calculate the overall mean, standard
deviation, and number of time slots from the result of the subquery. The ï¬nal result is a table
with 1 row and 3 columns.
12.5.2 Number of fully flagged baselines per antenna
The example below counts per antenna the number of fully flagged baselines, excluding the autocorrelations.
It uses grouping and aggregate functions twice; ï¬rst per baseline, thereafter per antenna. It also uses the
concatenation and backreferencing features. It also shows the results when using the time keyword, which
shows that the processing time is dominated by the ï¬rst query.
time select gsum(t1.cnt), t1.ANTENNA,
        (select NAME from ~/data/3C343.MS::ANTENNA)[t1.ANTENNA] as NAME
   from (select gcount() as cnt, ANTENNA1 as ANTENNA, ANTENNA2
           from ~/data/3C343.MS where all(FLAG) and ANTENNA1!=ANTENNA2
           groupby ANTENNA1,ANTENNA2) t0,
   [t0,
    (select cnt, ANTENNA2 as ANTENNA, 0 as ANTENNA2 from t0),
    (select 0 as cnt, rowid() as ANTENNA, 0 as ANTENNA2
            from ~/data/3C343.MS::ANTENNA)] t1
   groupby t1.ANTENNA orderby t1.ANTENNA
Â
   From query          0.8 real        0.58 user        0.08 system
   From query            0 real           0 user           0 system
   From query            0 real           0 user           0 system
   Subquery              0 real           0 user           0 system
   Groupby               0 real           0 user           0 system
   Orderby               0 real           0 user           0 system
   Projection            0 real           0 user           0 system
  Total time          0.83 real         0.6 user        0.08 system
The ï¬rst inner select counts the number of fully flagged baselines per baseline and stores the
result with the antennas making up a baseline in a temporary table with shorthand t0.. In this
way the expensive counting part needs to be executed only once.
This table has to be summed for both antennas of the baselines, which is done in the second
part creating a concatenated table with shorthand t1. It backreferences the ï¬rst table t0 twice.
First to use it directly to count for ANTENNA1, thereafter to count for ANTENNA2 by doing
a select to make the ANTENNA2 the ANTENNA column.
The table does not contain the antennas having no fully flagged baselines. Therefore the third
part of the concatenation copies the ANTENNA table to insert zero counts for all antennas.
Finally the outer select (in the ï¬rst line) sums the values in the concatenated table per antenna.
It also retrieves the name of the antenna by indexing in the selection of all antenna names.
13 Interface to TaQL
User and a programmer interfaces to TaQL are available. The program taql and some Python and
Glish functions form the user interface, while C++ classes and functions form the programmer
interface.
13.1 Python interface python-casacore
The main TaQL interface in Python is formed by the query function in module table. The function can be
used to compose and execute a TaQL command using the various (optional) arguments given to the query
function. E.g.
   import casacore.tables as pt
    tab = pt.table(’mytable’)
    seltab1 = tab.query (’column1 > 0’)
    seltab2 = seltab1.query (query=’column2>5’,
                             sortlist=’time’,
                             columns=’column1,column2’,
                             name=’result.tab’)
The ï¬rst command opens the table mytable. The second command does a simple query resulting in a
temporary table. That temporary table is used in the next command resulting in a persistent table. The
latter function call is transformed to the TaQL command:
  SELECT column1,column2 FROM \$1 WHERE column2>5
   ORDERBY time GIVING result.tab
During execution $1 is replaced by table seltab1. Note that the name argument generates the GIVING part to make the result persistent.
The functions sort and select exist as convenience functions for a query consisting of a sort or
column selection only. Both functions have an optional second name parameter to make the result
persistent.
   t1 = tab.sort (’time’)
    t1 = tab.select (’column1,column2’)
The calc function can be used to execute a TaQL calc command on the current table. The result can be
kept in a variable. For example, the following returns a vector containing the median of the DATA column in
each table row:
  med = t.calc (’median(DATA)’)
It is possible to embed Python variables and expressions in a TaQL command using the syntax
$variable and $(expression). A variable can be a standard numeric or string scalar or vector. It
can also be a table tool. An expression has to result in a numeric or string scalar or vector.
E.g
  from casacore.tables import *
   tab = table(’mytable’)
   coldata = tab.getcol (’col’);
   colmean = sum(coldata) / len(coldata);
   seltab1 = tab.query (’col > $colmean’)
   seltab2 = tab.query (’col > $(sum(coldata)/len(coldata))’)
   seltab3 = tab.query (’col > mean([SELECT col from $tab])’)
These three queries give the same result. The substitution mechanism is described in more detail in pyrap.util.
The most generic function that can be used is taql (or its synonym tablecommand). The full TaQL
command has to be given to that command. The result is a table object. E.g.
  import pyrap.tables as pt
   t = pt.taql(’select from GER.MS where ANTENNA1==1’);
By default, these commands will use the Python style for a TaQL statement. The style argument can be
used to choose another style.
13.2 Interface to Glish
The Glish interface is formed by script table.g. By default, it will use the Glish style for a TaQL statement.
For example:
  include ’table.g’
   tab := table(’mytable’)
   seltab1 := tab.query (’column1 > 0’)
   seltab2 := seltab1.query (query=’column2>5’,
                             sortlist=’time’,
                             columns=’column1,column2’,
                             name=’result.tab’)
   t := tablecommand(’select from GER.MS where ANTENNA1==1’,
                     style=’’);    # use default (glish) style
   med := t.calc (’median(DATA)’)
13.3 Program taql
The program taql makes it possible to execute TaQL commands from the shell. Commands can be given in
different ways:
The TaQL command can be given directly as command line arguments to the taql program.
The arguments will be combined to a single command (separated by a space). Note that using
multiple arguments instead of a single (quoted) argument makes it easier to use tab-completion
for the table name. It will execute the command, show the result, and exit.
Using the -f option, the name of a ï¬le containing one or more TaQL commands can be given. The
commands can be split over multiple lines, where a # can be used for comments. A semicolon
has to be used to separate commands. The special commands listed below for the interactive
case, can also be given.
The program is run interactively if command nor -f is given. It will run until the user stops via the
command exit, quit, or q or by giving Ctrl/D. Command editing and recall is possible, unless taql was
built without readline support. Of course, a ï¬le can be used as input by redirection to stdin.
This is more or less the same as using -f, but a command cannot be split over multiple
lines and no semicolon is needed to separate commands. Interactive commands are kept
in $HOME/.taql_history making command recall possible across multiple taql sessions.
A few types of commands can be given:
help, h or ? shows brief help information.
show can be used to show to available units or measure types.
A full TaQL command. If no GIVING part is given, the resulting table is not made
persistent. It shows the number of matching rows.
A full TaQL command preceded by varname=, where varname is the name under which
the resulting table is kept in this session. Thus the result is not a persistent table (unless
GIVING was given), but it is kept temporarily. The varname can be used in subsequent
commands like SELECT FROM $varname.
varname= without a further command removes the temporary result.
varname shows the number of rows in the temporary result. It can be followed by one or
more question marks to show the column names and details about them. Note that if an unknown varname is given, it is treated as a TaQL command resulting in a
parse error.
Comments can be given after the hash (#) character. Empty lines are ignored.
If columns are selected in the TaQL command, their values are printed. By default they are separated by tabs,
but the -d option can be used to deï¬ne another delimiter. Epoch, position, and direction measures
are printed in a formatted way. Note that TaQL’s str function offers nice formatting for any
value. Also the result of CALC commands is printed. Otherwise only the number of selected (or
updated or deleted) rows is printed. Options like -nops can be given to suppress printing the
results.
By default the program taql uses the Python style. The -s option can be used to specify a
style.
13.4 C++ interface
The C++ programmer can use TaQL commands and expressions at various levels,
13.4.1 TaQL query string
The function tableCommand in TableParse.h can be used to execute a TaQL command. The result is a Table
object. E.g.,
  Table seltab1 = tableCommand
          ("select from mytable where column1>0");
   Table seltab2 = tableCommand
          ("select column1,column2 from $1 where column2>5"
           " orderby time giving result.tab", seltab1);
These examples do the same as the Python ones shown above. Note that in the second function call the table name $1 is replaced by the object seltab1 passed to the
function. There is no style argument, so if an explicit style is needed it should be the ï¬rst part of the TaQL statement.
Note that the Glish style is the default style.
13.4.2 Expression string
The function parse in RecordGram.h can be used to parse a TaQL expression. The result is a
TableExprNode object that can be evaluated for each row in the table. E.g.
  Table tab("mytable");
   TableExprNode expr = RecordGram::parse (tab, "column1>0");
   Table seltab1 = tab(expr);
The example above does the same as the ï¬rst example in the previous section. There are, however, better
ways to use this functionality.
  Table tab("somename");
   TableExprNode expr = RecordGram::parse (tab, "ANTENNA1=1");
   for (uInt row=0; row<tab.nrow(); ++row) {
     if (expr.getBool(row)) {
       // expression is true for this row, so do something ...
     }
   }
The example above shows a boolean scalar expression, but it can also be a numeric expression or an array
expression as shown in the example below. Note that TaQL expression results have data type Bool, Int64,
Double, DComplex, String, or MVTime.
  TableExprNode expr = RecordGram::parse (tab, "abs(DATA)");
   Array<Double> data;
   for (uInt row=0; row<tab.nrow(); ++row) {
     expr.get (row, data);
   }
Class RecordGram can also be used to apply TaQL to C++ vectors of values or Records. The
RecordGram class documentation and its test program describe these features in more detail.
13.4.3 Expression classes
The other expression interface is a true C++ interface having the advantage that C++ variables can be used
directly. Class Table contains functions to sort a table or to select columns or rows. When selecting rows class
TableExprNode (in ExprNode.h) has to be used to build a WHERE expression which can be executed by the
overloaded function operator in class Table. E.g.
  Int limit = 0;
   Table tab ("mytable");
   Table seltab = tab(tab.col("column1") > limit);
does the same as the ï¬rst example shown above. See classes Table, TableExprNode, and TableExprNodeSet
for more information on how to construct a WHERE expression.
14 Writing user deï¬ned functions
A C++ user deï¬ned function has to be written as a class derived from the abstract base class
UDFBase. The documentation of this base class describes how to write a UDF. Furthermore
one can look at class UDFMSCal that contains the UDFs described in subsection User deï¬ned
functions.
It is possible to write a UDF that operates on an individual expression (for each table row) and returns
the result. It is, however, also possible to write a UDF acting as an aggregate function. In that case it will
return a result based on the values of all rows in a group. See the desription of the GROUPBY clause for
more information on the GROUPBY clause and aggregate functions.
Note that a class can contain multiple UDFs as done in UDFMSCal. Also note that a single UDF can
operate on multiple data types which is similar to a function like min that can operate on scalars and arrays
of different data types.
A UDF class can contain a HELP function, which should return help information. This function is called
by a help command like
   show function meas [subtype]
It returns an overview of the functions in the UDF class and possible other information. The optional
subtype argument can be used to return more speciï¬c information. Note that the same result is given
by
   meas.help(’subtype’)
TaQL ï¬nds a UDF by looking in a dictionary mapping the UDF name to a function constructing an
object of the UDF class. If not found, it tries to load the shared library with the lowercase name of the
library part of the UDF (like in derivedmscal.pa1). If the load is successful, it calls an initialization
function in the shared library that should add all UDF functions in the library to the dictionary. The
description of the UDFBase class shows how this should be done.
14.1 UDFs in Python
NOTE: This section is for a future version of TaQL. It has not been fully implemented yet.
For performance reasons User Deï¬ned Functions will usually be implemented in C++. It is, however,
possible to implement them in Python, both regular functions and aggregate functions. This can be done by
means of the pytaql module of Casacore.
A UDF has to be implemented in Python by subclassing pytaqlbase, that can be imported from
Casacore.python. The subclass has to implement a few functions, some are optional. The functions are
called in the order given below.
__init__(self) The class constructor must call the __init__ function of the superclass.
needTable(self) This optional function tells if the UDF needs the Table object of the table being queried. If
True is returned, the function setTable is called thereafter. The Table object can be used by
UDFs needing extra info (e.g., keywords) from the table being queried or from its subtables
(comparable to derivedmscal). It requires the import of pyrap.tables at the beginning of the
UDF script.
setTable(self, tab) This function makes it possible to keep the Table object. It must be implemented if needTable returns
True. The object should be kept like:
    self.tab = pyrap.tables.table (tab, _oper=3)
setup(self, valuetypes, datatypes, units) This function is called once at the beginning. It gets the value types, data types, and units of the
function arguments. The length of the sequences tells the number of arguments.
 valuetypes  int seq   value type of each argument
                        0=scalar 1=array 2=set
  datatypes   int seq   data type of each argument
                        0=bool 1=int 2=float 3=complex
                        4=string 6=date
  units       strings   unit of each argument (empty=no unit)
The UDF should check if the argument types are correct and determine the result type. It has to
return a dict containing the following ï¬elds:
 ndim      int       dimensionality of result
                      -1=scalar 0=unknown
  shape     int seq   shape (if fixed, otherwise empty sequence)
  dtype     int       data type of result
  unit      string    optional unit of result
  isaggr    bool      True = UDF is aggregate function
getValue(self, argValues, rownr) This function must return the function value for the given argument values. It is only called if setup
does not set isaggr=True. Normally the rownr argument is not needed, but it could be used to obtain
special info from the Table object for that row.
getAggrValue(self, rownrs) This function must return the value of the aggregate function for the given rows. The argument values
are not passed, because their sizes may exhaust memory. Instead the list of row numbers in the
aggregation group are given. For each row the following function must be called to get a list of the
argument values for that row.
    getArgValues(self, rownr)
Such a UDF can be called in TaQL like py.module.class where the class defaults to the module name.
An example of UDFs in Python is given below. The ï¬rst one is a regular UDF, the second one an aggregate
UDF.
# tpytaql.py: Test script for pytaqlbase
Â
 from casacore.pytaqlbase import pytaqlbase
 ##import pyrap.tables as pt
Â
 class tpytaql(pytaqlbase):
     """
     A test (and example) for a pytaql UDF.
     It returns the sum of the values in the argument.
     """
Â
     def __init__(self):
         """ The constructor must call the __init__ in the base class. """
         pytaqlbase.__init__(self)
Â
     def setup(self, valuetypes, datatypes, units):
         """ Setup the pytaql object. """
         if len(valuetypes) != 1:
             raise ValueError("UDF tpytaql should have exactly 1 argument")
         self.isScalar = valuetypes[0]==0
         return {’ndim’:0, ’dtype’:datatypes[0], ’unit’:units[0]}
Â
     def getValue(self, argValues, rownr):
         """ Get the value of the function for the given table row. """
         if self.isScalar:
             return argValues[0];
         return argValues[0].sum()    # sum of numpy array
Â
Â
Â
 class tpytaqlaggr(pytaqlbase):
     """
     A test (and example) for a pytaql aggregation UDF.
     It returns the difference of the total of both arguments.
     """
Â
     def __init__(self):
         """ The constructor must call the __init__ in the base class. """
         pytaqlbase.__init__(self)
Â
 ## The following functions show how to get and keep a Table object.
 ## Note the import of pyrap.tables is also required.
 ##    def needTable(self):
 ##        return True
 ##    def setTable(self, tab):
 ##        self.tab = pt.table(tab, _oper=3)
 ##        print "nrows",self.tab.nrows()
Â
     def setup(self, valuetypes, datatypes, units):
         """ Setup the pytaql object. """
         if len(valuetypes) != 2:
             raise ValueError("tpytaqlaggr UDF should have exactly 2 arguments")
         self.isScalar0 = valuetypes[0]==0
         self.isScalar1 = valuetypes[1]==0
         return {’ndim’:0, ’dtype’:datatypes[0],
                 ’unit’:units[0], ’isaggr’:True}
Â
     def getAggrValue(self, rownrs):
         """ Get the value of the aggregate function for the given table rows. """
         v = 0;
         for rownr in rownrs:
             argValues = self.getArgValues (rownr);
             if self.isScalar0:
                 v += argValues[0];
             else:
                 v += argValues[0].sum()
             if self.isScalar1:
                 v -= argValues[1];
             else:
                 v -= argValues[1].sum()
         return v
15 Possible future developments
In the near or far future TaQL can be enhanced by adding new features and by doing optimizations.
Add ROLLUP/CUBE to the GROUPBY clause.
Implement the OVER/PARTITION clause.
Add explicit JOIN clause (probably only equi-joins).
Add UNION, INTERSECTION, and DIFFERENCE.
Handle invalid subexpressions (e.g., exceeding array bounds) as null arrays which can be tested
with the function ISNULL.
pyrap is a Python binding to casacore classes using Boost.Python.It consists of a set of standard converters and bindings to the classes.As much as possible the bindings are the same as in glish.
Since long glish bindings to the casacore system have been in place. Quite
recently Python bindings have been created in the general casapy framework
using tools like CCMTools, Xerces, Xalan, and IDL. Albeit very flexible, it is
quite complicated and it is not straightforward to build on other systems than
RedHat and OS-X.
Therefore an attempt has been made to make a simpler Python binding using
Boost.Python. This proved to be very easy and succesful. The binding consists of
two parts:
Converters to translate objects between Python and C++.
Class wrappers to map a C++ class and its functions to Python.
The Python numarray and numpy (version 1.0 or higher) packages are supported. At
build time one can choose which ones should be used.
2 Converters
Boost.Python offers a nice way to convert objects to and from Python. Ralf W.
Grosse-Kunstleve <rwgk@yahoo.com> of Lawrence Berkeley National Laboratory
has built converters for standard STL containers. This has been extended to
convert to/from other objects.
The following C++ objects are currently supported:
scalars (bool, integer, real, complex)
std::string
casa::String
std::vector<T>
casa::Vector<T>
casa::IPosition
casa::Record
casa::ValueHolder
exceptions (casa::IterError and std::exception)
These C++ objects can usually be created from several types of Python objects and
are converted to a specific Python object.
A vector or IPosition object is converted to a Python list.
It can be constructed from the following Python objects:
scalar
list or tuple
numarray scalar or 1-dim array
numpy scalar or 1-dim array
Note that a list or tuple of arbitrary objects can be given. For example, it is
possible to get a Vector<TableProxy> from Python.
A casa::Record is mapped to a Python dict.
Every C++ exception is mapped to a Python RunTimeError exception.
However, casa::IterError is special and is mapped to an end-of-iteration
exception (StopIteration) in Python.
A casa::ValueHolder is a special casacore object that can hold a record
or a scalar value or n-dim array of many types (bool, numeric,
string). It is meant to conceal the actual type which is useful in
functions that can accept a variety of types (like getcell in the table
binding).
Converting a ValueHolder to Python creates the appropriate Python scalar,
array, or dict object. When converting from Python to ValueHolder,
the appropriate internal ValueHolder value is constructed; a list,
tuple, and array object are converted to an casacore array in the
ValueHolder.
It means there is no direct Array conversion to/from Python. A ValueHolder
object is always needed to do the conversion. Note that this is a cheap
operation, as it uses Array reference semantics. ValueHolder has functions to
convert between types, so one can get out an Array with the required
type.
2.1 Array conversion to/from numpy and numarray
casacore arrays are kept in Fortran-order, while Python arrays are kept in
C-order. It was felt that the Python interface should be as pythonic as possible.
Therefore it was decided that the array axes are reversed when converting
to/from Python. The values in an IPosition object (describing shape or position)
are also reversed when converting to/from Python.
Note that although numarray and numpy have Fortran-array provisions by
setting the appropriate internal strides, they do not really support them. When
adding, for instance, the scalar value 0 to a Fortran-array, the result
is a transposed version of the original (which can be a quite expensive
operation).
A function binding could be such that shape information is passed via, say, a
Record and not via an IPosition object. In that case its values are
not reversed automatically, so the programmer is responsible for doing
it.
An casacore array is returned to Python as an array object containing a copy
of the casacore array data. If pyrap has been built with support for only one
Python array package (numpy or numarray), it is clear which array type is
returned. If support for both packages has been built in, by default an array of
the imported package is returned. If both or no array packages have been
imported, a numpy array is returned.
Note that there is no support for the old Numeric package.
An casacore array constructed from a Python array is regarded as a
temporary object. So if possible, the casacore array refers to the Python array
data to avoid a needless copy. This is not possible if the element size in Python
differs from casacore. It is also not possible if the Python array is not
contiguous (or not aligned or byte swapped). In those cases a copy is
made.
A few more numarray/numpy specific issues are dealt with:
An empty N-dim casacore array (i.e. an array containing no elements)
is returned as an empty N-dim Python array. If the dimensionality
is zero, it is returned as an empty 1-dim array, to prevent
numarray/numpy from treating it as a scalar value.
In numarray array() results in Py_None. This is accepted by the
converters as an empty 1-dim array.
Empty arrays can be constructed in Python using empty lists. For
example, array([[]]) results in an empty 2-dim array. The converters
accept such empty N-dim Python arrays. The type of an empty array
is set to Int by numarray and to Double by numpy.
Because the type of an empty Python array cannot easily be set, the
converters can convert an empty integer or real array to any type.
The converters accept a numpy string array. However, it is returned
to Python as the special dict object described above.
3 Class wrappers
Usually a binding to an existing Proxy class is made, for example TableProxy,
which should be the same class used in the glish-binding. For a simple binding,
only some simple C++ code has to be written in pyrap_xx/src/pyxx.cc, where
XX is the name of the package/class.
// Include files for converters being used.
#include <casacore/python/Converters/PycExcp.h>
#include <casacore/python/Converters/PycBasicData.h>
#include <casacore/python/Converters/PycRecord.h>
// Include file for boost python.
#include <boost/python.hpp>
using namespace boost::python;
namespace casa { namespace pyrap {
void wrap_xx()
{
// Define the class; "xx" is the class name in Python.
class_<XX> ("xx")
// Define the constructor.
// Multiple constructors can be defined.
// They have to have different number of arguments.
.def (init<>())
// Add a .def line for each function to be wrapped.
// An arg line should be added for each argument giving
// its name and possibly default value.
.def ("func1", &XX::func1,
(boost::python::arg("arg1"),
boost::python::arg("arg2")=0))
;
}
}}
BOOST_PYTHON_MODULE(_xx)
{
// Register the conversion functions.
casa::pyrap::register_convert_excp();
casa::pyrap::register_convert_basicdata();
casa::pyrap::register_convert_casa_record();
// Initialize the wrapping.
casa::pyrap::wrap_xx();
}
Python requires for each package a file __init__.py, so such an empty file
should be created as well.
3.1 More complicated wrappers
Sometimes a C++ function cannot be wrapped directly, because the argument
order needs to be changed or because some extra Python checks are
necessary. In such a case the class needs to be implemented in Python
itself.
The C++ wrapped class name needs to get a different name, usually by
preceeding it with an underscore like:
class_<XX> ("_xx")
The Python class should be derived from it and implement the constructor by
calling the constructor of _xx.
class xx(_xx):
def __init__(self):
_xx.__init__(self)
Now xx inherits all functions from _xx. The required function can be written in
Python like
Note that in the wrapper the function name also needs to be preceeded by an
underscore to make it different.
3.2 Combining multiple classes
Sometimes one wants to combine multiple classes in a package. A example is
package pyrap_tables which contains the classes table, tablecolumn,
tablerow, tableiter, and tableindex. One is referred to the code of this
package to see how to do it.
4 Python specifics
Besides an array being in C-order, there are a few more Python specific
issues.
Indexing starts at 0 (vs. 1 in glish).
The end value in a range like [10:20] is exclusive (vs. inclusive in
glish). Furthermore Python supports a step and reversed ranges.
Where useful, the function __str__ should be added giving the name
of the object. This function is used when printing an object.
Where useful, the functions __len__, __setitem__(index, value),
and __getitem__(index) should be added to make it possible
that a user indexes an object directly like tabcol[i] or
tabcol[start:stop:step].
When these functions are added, Python supports iteration in
an object. Explicit iteration can also be done by adding the
functions __iter__ and next. At the end next should raise the
Python StopIteration exception (or throw casa::IterError when
implemented in C++) to stop the iteration.